Open Positions – ERTS Lab

Level: Master

Motivation

A previous feasibility study successfully demonstrated that an IMU-based system can capture movement patterns of parrots at Wuppertal Zoo without fixed infrastructure in the enclosure. The prototype — built around an nRF54L15-DK with an LSM6DS3, LIS3MDL, and BMP280 — showed that heading estimation, altitude detection, and rotation-rich trajectory tracking are achievable in principle.

However, the study also exposed fundamental limitations: dead-reckoning drift accumulates over time, uniform straight-line motion cannot be detected, and the velocity estimation relies on a calibrated fallback radius that introduces systematic errors for non-circular motion. Furthermore, the state of the art in inertial localisation and sensor fusion was only superficially covered — a gap this project aims to close systematically.

The central question for this follow-up project is therefore: given the current state of research and technology, what is the best architecture for a practical, wearable localisation system for small animals?

Objectives

The project pursues three interconnected goals:

• A comprehensive review and comparison of the state of the art in inertial localisation, sensor fusion, and activity recognition for wearable systems, with particular attention to approaches relevant for small, unconstrained animals
• A technology and algorithm evaluation that leads to a well-founded recommendation for hardware, sensor selection, and fusion strategy for this use case
• A proof-of-concept implementation of the recommended approach, validated against the baseline established by the feasibility study

Starting Point

The existing system and its documented results serve as the baseline for this project. Key findings from the feasibility study that directly shape the direction of this work:

• Dead reckoning alone is insufficient for absolute positional accuracy — drift is a fundamental property, not an implementation flaw
• Straight-line motion at constant velocity is undetectable with the rotation-rate-based velocity estimator; the system relies entirely on changes in direction or speed
• The centripetal velocity term v = a⊥/ω is rendered unusable by the required high-pass filter, forcing a fixed calibrated fallback radius
• The BMP280 altitude measurement is functional but noisy; the study explicitly recommends upgrading to a BMP388
• The nRF54L15-DK development board is too large for attachment to an animal — a compact PCB is needed for real-world trials
• The existing UWB infrastructure at Wuppertal Zoo represents an available external reference that could be used to periodically correct accumulated drift

Planned Approach

1. State-of-the-art review

- Survey inertial navigation and dead-reckoning systems for wearable and animal-tracking applications

- Compare filter and fusion architectures: complementary filters, Madgwick/Mahony, Extended Kalman Filter (EKF), Error-State Kalman Filter (ESKF), and learning-based approaches (e.g. RoNIN, IONet)

- Review sensor fusion strategies combining IMU with barometry, magnetometry, UWB, and BLE

- Identify ZUPT variants and step-detection methods suited to non-pedestrian motion profiles

- Evaluate hardware options: IMU sensors, magnetometers, pressure sensors, and MCU platforms relevant for ultra-low-power wearable designs

2. Technology and algorithm evaluation

- Assess which sensor fusion approach best matches the constraints of this use case: irregular motion, small device size, battery operation, no permanent external reference

- Compare IMU-only fusion against IMU+UWB and IMU+BLE hybrid approaches, considering the available zoo infrastructure

- Evaluate whether MEMS-grade sensors are sufficient or whether higher-grade alternatives are warranted

- Produce a documented recommendation for the target architecture

3. Implementation and validation

- Implement the recommended fusion algorithm (e.g. ESKF or Madgwick with UWB correction) on the existing or updated hardware

- Compare performance against the baseline established in the feasibility study using identical test scenarios (circular path, zigzag, straight-line, altitude step)

- Evaluate heading accuracy, trajectory drift, distance error, and altitude resolution

Research Questions

• What fusion architectures from current literature are most applicable to unconstrained animal motion, and what are their practical trade-offs?
• Does integrating UWB position fixes from the zoo's existing infrastructure sufficiently bound dead-reckoning drift for behavioural research purposes?
• Can BLE-based ranging or fingerprinting serve as a lightweight alternative or complement to UWB for drift correction?
• What IMU, magnetometer, and pressure sensor combination best balances accuracy, power consumption, and physical size for a wearable animal tracker?
• Does a more sophisticated fusion algorithm (e.g. ESKF) yield a meaningful improvement in trajectory accuracy over the calibrated fallback approach from the feasibility study, given the same sensor hardware?
• What are the minimum hardware requirements — in terms of sensor noise floor, sampling rate, and processing power — to make the system viable for real animal trials?

Expected Contribution

The project delivers:

• A structured state-of-the-art review of inertial and sensor-fusion localisation relevant to wearable animal tracking, filling the gap identified in the feasibility study
• A documented technology decision — hardware selection and fusion architecture — with explicit justification based on the review and the constraints of the use case
• A validated prototype implementation of the recommended approach, benchmarked against the existing baseline
• A concrete foundation for subsequent hardware miniaturisation and real-world animal trials at Wuppertal Zoo

The project bridges the gap between a demonstrated proof of concept and a system design that is ready for further engineering toward practical deployment.

Contact: Markus Cremer ([email protected])

Level: Bachelor or Master

Motivation

Evaluating localisation systems — whether based on BLE, UWB, RFID, or other technologies — requires systematic measurements at well-defined positions in space. The current process relies on manually placing a tag or sensor node on a 2D grid (50×50 cm grid points, PVC floor, wooden stand at fixed height), followed by an is/should comparison to determine the mean localisation error. A motorised linear trolley (stepper motor on a wooden rail) exists for channel measurements along a single axis.

This manual approach is time-consuming, limits reproducibility, and is restricted to two dimensions. Extending measurements to 3D and automating the positioning process would significantly increase both the throughput and the scientific rigour of localisation experiments.

Objectives

The goal of this project is to develop an automated system capable of positioning a measurement point — tag, antenna, or sensor — reproducibly in three dimensions. The system must be scalable to room size but will initially be realised as a table-scale prototype. Key requirements are positional accuracy, low RF interference introduced by the mechanical structure, and practical scalability.

Starting Point

Several positioning concepts have been identified as candidates for further investigation:

• Cartesian rail system (e.g. FarmBot-style gantry)
• Mobile ground robot (Roomba-style autonomous platform)
• Cable-suspended system (SpiderCam-style cable-driven parallel robot, CDPR)

Additional concepts such as delta robots or other cable robot variants are to be considered in an open literature and technology search. Based on an initial assessment, the cable-driven parallel robot (CDPR) is considered the most promising candidate due to its good scalability, minimal material presence in the measurement volume, and RF transparency.

Planned Approach

1. Concept analysis and comparison

- Research and document relevant positioning concepts including those beyond the initial candidate list

- Evaluate each concept against defined criteria: positional accuracy, mechanical complexity, scalability, cost, and RF interference with measurements

- Select the most promising concept with documented justification

2. Prototype development

- Design and build a table-scale prototype of the selected concept

- Implement motion control (stepper motors or equivalent) with defined positioning resolution

- Integrate a simple interface for specifying target positions and executing measurement sequences

3. Validation

- Characterise positional accuracy and repeatability of the prototype

- Assess RF influence of the mechanical structure on a representative localisation measurement

4. Preparation for scaling

- Document design decisions and identify what changes are required for scaling from table to room size

- Outline a concept for room-scale deployment

Research Questions

• Which positioning concept best satisfies the combined requirements of accuracy, scalability, low RF interference, and reasonable implementation effort?
• Can a cable-driven parallel robot (CDPR) achieve sufficient positional accuracy and repeatability for localisation system evaluation at table scale?
• What mechanical and control design choices are critical for later scaling to room size?
• How significantly does the mechanical structure of each candidate concept affect RF measurements in the relevant frequency bands?

Expected Contribution

The project delivers:

• A documented concept comparison covering all relevant 3D positioning approaches for localisation measurement automation
• A functional table-scale prototype of the selected concept, with motion control and a basic positioning interface
• Characterisation data on positional accuracy, repeatability, and RF influence
• A scaling concept that describes the path from table prototype to room-scale deployment for future localisation experiments

Contact: Markus Cremer ([email protected])

Level: Master (Final Thesis)

Motivation

Tracking assets or animals over large outdoor areas requires two things that are rarely available from a single compact, low-power device: global connectivity and reliable positioning. Conventional approaches either combine separate GNSS and cellular modules at high power cost, or rely on area-specific infrastructure such as LoRaWAN, which lacks global reach. Recent ultra-compact IoT modules such as the ST87M01 from STMicroelectronics change this equation: a single 10.6 × 12.8 mm module integrates NB-IoT (LTE CatNB2 Release 15), GNSS positioning, and Wi-Fi positioning, with a sleep current below 1.2 µA.

The availability of three positioning technologies on one platform introduces a fundamental design question: which technology should be used, when, and in what combination — given that each involves a very different trade-off between energy consumption, latency, and positional accuracy? GNSS provides metre-level accuracy but is slow and power-hungry to acquire; NB-IoT Cell-ID positioning is fast and cheap but coarse; Wi-Fi positioning bridges the gap in urban environments. Choosing blindly is wasteful; choosing adaptively requires a model of both the environment and the application's accuracy and reporting requirements.

This thesis addresses exactly that problem. It develops a custom hardware platform based on the ST87M01, characterises all three positioning modalities under realistic conditions, and derives a context-aware, energy-optimal technology selection strategy — providing the foundation for a practical, long-life asset or animal tracker with global connectivity.

Objectives

• A custom hardware design based on the ST87M01, dimensioned for battery-powered, wearable or attachable deployment
• A systematic characterisation of GNSS, NB-IoT Cell-ID, and Wi-Fi positioning in terms of accuracy, time-to-first-fix, and energy per position fix across representative outdoor and urban scenarios
• An energy model of the complete system covering all relevant operating states: GNSS acquisition, NB-IoT data transmission, Wi-Fi scan, MCU active and sleep modes, and IMU-based wake-up
• A context-aware positioning strategy — implemented and validated on the custom hardware — that selects and combines positioning technologies based on movement state, environment, and application-defined accuracy and update-rate requirements

Starting Point

The ST87M01 was released in late 2025 and represents the first commercially available module to integrate NB-IoT, GNSS, and Wi-Fi positioning in this form factor. Published characterisation data and independent evaluations of the module are limited, making systematic empirical work both timely and necessary.

The ERTS Lab provides relevant context for this thesis:

• Ongoing research in localisation systems (UWB, BLE, IMU-based), with established methodology for accuracy benchmarking and energy profiling
• An existing animal-tracking research thread at Wuppertal Zoo, which defines a concrete and demanding deployment scenario for the resulting system
• Expertise in ultra-low-power embedded design on Nordic and STM32 platforms, relevant to the firmware and power architecture of the tracker

ST provides an evaluation board (STEVAL-NBIOTV1) with the ST87M01, an STM32U5 host MCU, environmental sensors, and a Li-ion charger. This board serves as the initial bring-up and reference platform; the custom PCB developed in this thesis is designed for deployment rather than evaluation.

Planned Approach

1. Literature and technology review

- Review low-power GNSS strategies: cold start vs. assisted GNSS (A-GNSS via NB-IoT), snapshot GNSS, and duty-cycled acquisition

- Survey NB-IoT Cell-ID and OTDOA positioning: accuracy expectations and network dependency

- Review Wi-Fi positioning approaches available without dedicated infrastructure

- Identify relevant energy modelling approaches for multi-modal IoT systems

2. Custom hardware design

- Design a compact PCB integrating the ST87M01, a host MCU, LiPo battery management, an IMU for motion detection, and in-circuit current measurement shunts

- Target form factor and weight suitable for animal-borne or asset-mounted deployment

- Implement a firmware baseline covering all three positioning modalities, NB-IoT data uplink, and deep-sleep power management

3. Characterisation

- Measure energy per position fix, time-to-first-fix, and positional accuracy for each modality individually across defined outdoor, urban, and transitional scenarios

- Characterise the interaction between NB-IoT connectivity and A-GNSS acquisition time

- Record complete current profiles for all relevant operating state transitions

4. Energy modelling and strategy design

- Build an analytical energy model parameterised by positioning modality, update rate, and environmental context

- Design a technology selection strategy — implemented as a state machine or lightweight decision model — that minimises energy consumption subject to configurable accuracy and reporting-rate constraints

- Implement and validate the strategy on the custom hardware against the characterisation baseline

5. Field validation

- Evaluate the complete system in representative outdoor scenarios (open sky, urban canyon, partial canopy cover)

- Report position accuracy, system lifetime estimate, and NB-IoT coverage reliability

- Compare adaptive strategy against fixed single-technology baselines

Research Questions

• What is the measured energy cost per position fix for GNSS, NB-IoT Cell-ID, and Wi-Fi positioning under realistic outdoor conditions, and how does this vary across scenarios?
• How significantly does A-GNSS via NB-IoT reduce time-to-first-fix and acquisition energy relative to a cold start, and under what network conditions does this benefit degrade?
• Can NB-IoT Cell-ID positioning provide sufficient accuracy for coarse-grained animal or asset tracking when GNSS acquisition is undesirable?
• What movement-state and environment indicators — detectable from an onboard IMU and network feedback — are sufficient to drive effective technology selection?
• What duty-cycle and technology-selection strategy minimises energy consumption while satisfying a defined accuracy and update-rate requirement?
• How does a custom miniaturised PCB based on the ST87M01 compare to the ST evaluation platform in terms of power consumption, and what design choices are most impactful?

Expected Contribution

• A custom hardware design — schematic, PCB layout, assembled prototype — based on the ST87M01, suitable as a foundation for subsequent animal-tracking or asset-tracking deployments
• A characterisation dataset covering all three positioning modalities of the ST87M01 under realistic outdoor conditions, filling a gap in the currently sparse published literature on this module
• A validated energy model of the complete system across all relevant operating states
• A context-aware technology selection strategy, implemented and benchmarked against fixed single-modality baselines
• A concrete starting point for subsequent work integrating this outdoor tracking capability with the lab's existing indoor localisation research — for example, as the outdoor complement to UWB- or BLE-based indoor positioning in seamless tracking scenarios

Contact: Markus Cremer ([email protected])

Level: Bachelor (Abschlussarbeit)

Motivation

Battery-powered tracking devices for animals or mobile assets must balance two competing demands: positional accuracy and update rate on one side, energy consumption and battery lifetime on the other. The ST87M01 from STMicroelectronics offers a compelling hardware basis for this problem — integrating NB-IoT (LTE CatNB2 Release 15) and GNSS in a 10.6 × 12.8 mm module with a sleep current below 1.2 µA. However, the energy cost of each operating state, and the interaction between GNSS acquisition strategy and NB-IoT connectivity, have not been systematically characterised for this module.

This thesis develops a custom tracker hardware based on the ST87M01, constructs a validated analytical energy model covering all relevant operating states, and derives a movement-adaptive duty-cycle algorithm that adjusts the positioning rate to the activity level of the tracked subject — providing a concrete, quantified foundation for designing long-life tracking systems with this module.

Objectives

• Design and manufacture a compact custom PCB based on the ST87M01 with integrated IMU, battery management, and current measurement
• Develop firmware implementing GNSS acquisition, NB-IoT uplink, deep-sleep management, and IMU-based motion detection
• Construct and validate an analytical energy model parameterised by operating state, duty cycle, and GNSS acquisition strategy
• Design and evaluate a movement-adaptive duty-cycle algorithm that reduces the positioning rate during inactivity and increases it during motion

Starting Point

ST provides the STEVAL-NBIOTV1 evaluation board — featuring the ST87M01, an STM32U5 host MCU, an IMU, environmental sensors, and a Li-ion charger — as a reference platform for initial bring-up. This thesis uses the evaluation board for early firmware validation only; the custom PCB is the actual target platform.

The ERTS Lab provides relevant prior work in IMU-based motion detection, power profiling methodology, and embedded systems design on STM32 and Nordic platforms. An ongoing animal-tracking research thread at Wuppertal Zoo provides a concrete application scenario that motivates and frames the design choices in this thesis.

Planned Approach

1. Reference platform evaluation

- Bring up the ST87M01 on the STEVAL-NBIOTV1

- Characterise current consumption per operating state: GNSS acquisition (cold start and A-GNSS), NB-IoT transmission, IMU active, and deep sleep

- Establish baseline measurements against which the custom board will be compared

2. Custom PCB design and firmware

- Design and manufacture a compact custom board: ST87M01, low-power MCU, IMU, LiPo charger, measurement shunt

- Implement duty-cycle firmware: sleep → motion detection → conditional wake → GNSS fix → NB-IoT uplink → sleep

- Validate board functionality and repeat characterisation measurements

3. Energy modelling

- Construct an analytical model of total energy consumption as a function of positioning rate, GNSS strategy, and NB-IoT transmission parameters

- Validate the model against measured current profiles across multiple duty-cycle configurations

- Use the model to derive battery lifetime predictions for representative scenarios

4. Movement-adaptive duty-cycle algorithm

- Design a state machine that distinguishes resting and active states using the onboard IMU

- Implement differentiated duty cycles: reduced positioning rate during inactivity, increased rate during motion

- Evaluate energy savings and positional coverage relative to a fixed-rate baseline

- Validate under controlled motion scenarios representative of the animal-tracking use case

Research Questions

• What is the measured energy cost of GNSS acquisition, NB-IoT transmission, and deep sleep on the custom ST87M01 platform, and how accurately does an analytical model predict total consumption?
• How significantly does A-GNSS via NB-IoT reduce acquisition energy and time-to-first-fix relative to a cold start, and under what conditions is this benefit relevant?
• What battery lifetime is achievable under realistic duty-cycle configurations, and which operating state dominates the energy budget?
• What IMU-based activity indicators are sufficient to reliably distinguish resting from active states in a small-animal tracking context?
• How much energy does a movement-adaptive duty-cycle strategy save relative to a fixed update rate, and what is the trade-off in positional coverage?

Expected Contribution

• A custom hardware design — schematic, PCB layout, assembled prototype — based on the ST87M01, sized and powered for animal-borne or asset-mounted deployment
• A validated analytical energy model covering all relevant operating states, with measured parameterisation on the custom board
• A movement-adaptive duty-cycle algorithm implemented in firmware and evaluated against a fixed-rate baseline
• Quantified battery lifetime estimates for representative deployment scenarios, providing a concrete design reference for follow-up work
• A hardware and firmware foundation directly usable as the starting point for the master-level thesis on multi-technology adaptive positioning

Contact: Markus Cremer ([email protected])

Details available on request. –

Contact: Leon Schex ([email protected])

Motivation

In many battery-powered systems, additional sensors such as IMUs, magnetometers, or barometers do not need to be continuously active. Instead, a very low-power motion detector can deliver a signal or interrupt to an nRF54L15 already operating in System ON Idle. The nRF54L15 then powers up additional sensors only on demand, reads them out, and shuts them down again.

Crucially, not only a low-power motion detector is required, but also a sensor supply that causes virtually no leakage current in the off state.

Objectives

The goal of this project is to develop a circuit that:

• provides an ultra-low-power motion signal to the nRF54L15,
• delivers an interrupt/trigger to the nRF54L15,
• subsequently switches on additional sensors selectively,
• and allows virtually no current flow when the additional sensors are powered down.

The ADXL367 with a typical wake-up mode current of 180 nA serves as the reference. The project investigates whether a solution can be found that undercuts this reference in terms of motion-path current consumption and/or cost.

Starting Point

In this project, the nRF54L15 is not powered off but remains in System ON Idle. The motion detector therefore does not serve to power up the microcontroller, but solely as an interrupt or trigger signal.

After the interrupt, additional sensors are to be powered on. Several architectures are conceivable:

• Supply directly via GPIO
• Supply via load switch
• Supply via discrete MOSFET
• Additional measures against back-feeding through I/O lines

It is particularly important that powered-down sensors are not inadvertently kept alive through SDA, SCL, INT, or other signal lines. The project therefore explicitly addresses leakage currents through supply rails, pull-ups, and I/O paths.

Planned Approach

1. Analyze the reference

- Evaluate the ADXL367 as a benchmark for current consumption and cost

- Define target values for the motion path and sensor power-down

2. Compare solution approaches

- Integrated low-power accelerometer

- Passive sensor with comparator/Schmitt trigger

- Hybrid solutions for motion detection and sensor switching

3. Design the circuit

- Develop the motion interrupt stage for the nRF54L15

- Design the power switching for IMU, magnetometer, and barometer

- Compare variants using GPIO, load switch, and MOSFET

4. Develop PCB and measure

- Design a demonstrator board

- Measure quiescent current, leakage current, and interrupt behavior

- Perform comparison against the ADXL367

Research Questions

• Which motion detection concepts are more power-efficient or lower cost than the ADXL367?
• Can a passive sensor with a comparator be a viable alternative to a MEMS accelerometer?
• How can additional sensors be powered down such that virtually no current flows?
• Is switching via GPIOs sufficient, or is a load switch/MOSFET technically more appropriate?
• What leakage currents arise through supply rails, pull-ups, ESD structures, and I/O lines?
• How robust is the solution with respect to false triggers, interference, and real motion events?

Expected Contribution

The expected outcome is a compact demonstrator comprising:

• an ultra-low-power motion interrupt stage for the nRF54L15,
• selective power switching of additional sensors,
• a low-leakage power-down architecture,
• a PCB design,
• and measurement data on quiescent current, leakage current, trigger behavior, and cost.

The project aims to identify the most suitable architecture for a real low-power system: a dedicated motion sensor, or a simpler, lower-cost external motion circuit with rigorous power gating.

Contact: Leon Schex ([email protected])

Motivation

A previous feasibility study successfully demonstrated that an IMU-based system can capture movement patterns of parrots at Wuppertal Zoo without fixed infrastructure in the enclosure. The prototype — built around an nRF54L15-DK with an LSM6DS3, LIS3MDL, and BMP280 — showed that heading estimation, altitude detection, and rotation-rich trajectory tracking are achievable in principle.

However, the study also exposed fundamental limitations: dead-reckoning drift accumulates over time, uniform straight-line motion cannot be detected, and the velocity estimation relies on a calibrated fallback radius that introduces systematic errors for non-circular motion. Furthermore, the state of the art in inertial localisation and sensor fusion was only superficially covered — a gap this project aims to close systematically.

The central question for this follow-up project is therefore: given the current state of research and technology, what is the best architecture for a practical, wearable localisation system for small animals?

Objectives

The project pursues three interconnected goals:

• A comprehensive review and comparison of the state of the art in inertial localisation, sensor fusion, and activity recognition for wearable systems, with particular attention to approaches relevant for small, unconstrained animals
• A technology and algorithm evaluation that leads to a well-founded recommendation for hardware, sensor selection, and fusion strategy for this use case
• A proof-of-concept implementation of the recommended approach, validated against the baseline established by the feasibility study

Starting Point

The existing system and its documented results serve as the baseline for this project. Key findings from the feasibility study that directly shape the direction of this work:

• Dead reckoning alone is insufficient for absolute positional accuracy — drift is a fundamental property, not an implementation flaw
• Straight-line motion at constant velocity is undetectable with the rotation-rate-based velocity estimator; the system relies entirely on changes in direction or speed
• The centripetal velocity term v = a⊥/ω is rendered unusable by the required high-pass filter, forcing a fixed calibrated fallback radius
• The BMP280 altitude measurement is functional but noisy; the study explicitly recommends upgrading to a BMP388
• The nRF54L15-DK development board is too large for attachment to an animal — a compact PCB is needed for real-world trials
• The existing UWB infrastructure at Wuppertal Zoo represents an available external reference that could be used to periodically correct accumulated drift

Planned Approach

1. State-of-the-art review

- Survey inertial navigation and dead-reckoning systems for wearable and animal-tracking applications

- Compare filter and fusion architectures: complementary filters, Madgwick/Mahony, Extended Kalman Filter (EKF), Error-State Kalman Filter (ESKF), and learning-based approaches (e.g. RoNIN, IONet)

- Review sensor fusion strategies combining IMU with barometry, magnetometry, UWB, and BLE

- Identify ZUPT variants and step-detection methods suited to non-pedestrian motion profiles

- Evaluate hardware options: IMU sensors, magnetometers, pressure sensors, and MCU platforms relevant for ultra-low-power wearable designs

2. Technology and algorithm evaluation

- Assess which sensor fusion approach best matches the constraints of this use case: irregular motion, small device size, battery operation, no permanent external reference

- Compare IMU-only fusion against IMU+UWB and IMU+BLE hybrid approaches, considering the available zoo infrastructure

- Evaluate whether MEMS-grade sensors are sufficient or whether higher-grade alternatives are warranted

- Produce a documented recommendation for the target architecture

3. Implementation and validation

- Implement the recommended fusion algorithm (e.g. ESKF or Madgwick with UWB correction) on the existing or updated hardware

- Compare performance against the baseline established in the feasibility study using identical test scenarios (circular path, zigzag, straight-line, altitude step)

- Evaluate heading accuracy, trajectory drift, distance error, and altitude resolution

Research Questions

• What fusion architectures from current literature are most applicable to unconstrained animal motion, and what are their practical trade-offs?
• Does integrating UWB position fixes from the zoo's existing infrastructure sufficiently bound dead-reckoning drift for behavioural research purposes?
• Can BLE-based ranging or fingerprinting serve as a lightweight alternative or complement to UWB for drift correction?
• What IMU, magnetometer, and pressure sensor combination best balances accuracy, power consumption, and physical size for a wearable animal tracker?
• Does a more sophisticated fusion algorithm (e.g. ESKF) yield a meaningful improvement in trajectory accuracy over the calibrated fallback approach from the feasibility study, given the same sensor hardware?
• What are the minimum hardware requirements — in terms of sensor noise floor, sampling rate, and processing power — to make the system viable for real animal trials?

Expected Contribution

The project delivers:

• A structured state-of-the-art review of inertial and sensor-fusion localisation relevant to wearable animal tracking, filling the gap identified in the feasibility study
• A documented technology decision — hardware selection and fusion architecture — with explicit justification based on the review and the constraints of the use case
• A validated prototype implementation of the recommended approach, benchmarked against the existing baseline
• A concrete foundation for subsequent hardware miniaturisation and real-world animal trials at Wuppertal Zoo

The project bridges the gap between a demonstrated proof of concept and a system design that is ready for further engineering toward practical deployment.

Contact: Markus Cremer ([email protected])

Level: Master (Research Project)

Motivation

The ST87M01 from STMicroelectronics is among the first commercially available modules to integrate NB-IoT (LTE CatNB2 Release 15), GNSS, and Wi-Fi positioning in a single 10.6 × 12.8 mm package with a sleep current below 1.2 µA. Despite its potential for low-power asset and animal tracking, systematic and independent characterisation data for this module is virtually absent from the literature.

This research project focuses on the empirical and analytical groundwork that any serious application of the ST87M01 requires: a rigorous characterisation of all three positioning modalities and a validated energy model of the complete system. The work is deliberately scoped to the ST evaluation platform — keeping hardware effort minimal — so that the full project duration can be invested in measurement methodology, model construction, and analysis. The results directly feed into the lab's ongoing animal-tracking research and serve as the foundation for the accompanying master's thesis.

Objectives

• Characterise all three positioning modalities of the ST87M01 — GNSS, NB-IoT Cell-ID, and Wi-Fi positioning — in terms of accuracy, time-to-first-fix, and energy per position fix across representative scenarios
• Construct and validate an analytical energy model of the complete system, covering all relevant operating states and their transitions
• Evaluate assisted GNSS (A-GNSS) via NB-IoT as a strategy for reducing acquisition energy and time-to-first-fix
• Derive design guidelines for duty-cycle configuration and technology selection in low-power tracking applications

Starting Point

ST provides the STEVAL-NBIOTV1 evaluation board — featuring the ST87M01, an STM32U5 host MCU, an IMU, environmental sensors, and a Li-ion charger — as the hardware platform for this project. No custom PCB is required; the evaluation board provides sufficient access for firmware development and current profiling.

The ERTS Lab provides established methodology for energy measurement and localisation benchmarking, as well as the application context from its ongoing animal-tracking work at Wuppertal Zoo.

Planned Approach

1. Literature review

- Survey low-power GNSS acquisition strategies: cold start, A-GNSS, snapshot GNSS

- Review NB-IoT Cell-ID and OTDOA positioning: accuracy and network dependency

- Review Wi-Fi positioning without dedicated infrastructure

- Identify energy modelling approaches for multi-modal IoT systems

2. Measurement setup and firmware

- Develop firmware on the STEVAL-NBIOTV1 to exercise each positioning modality independently and in combination

- Instrument the board for accurate current measurement across all operating states

- Define a structured measurement protocol covering all relevant scenario types

3. Characterisation measurements

- Measure energy per fix, time-to-first-fix, and positional accuracy for GNSS (cold start and A-GNSS), NB-IoT Cell-ID, and Wi-Fi positioning

- Cover at least three scenario types: open sky, urban canyon, and indoor/transitional

- Record full current profiles for all operating state transitions

4. Energy modelling

- Construct an analytical energy model parameterised by modality, duty cycle, and scenario

- Validate the model against measured profiles across multiple configurations

- Use the model to derive battery lifetime estimates and identify the dominant energy contributors

5. Design guidelines

- Derive concrete recommendations for duty-cycle configuration and technology selection as a function of accuracy requirement, update rate, and scenario

- Document findings as a reference for the accompanying master's thesis and future hardware designs

Research Questions

• What is the measured energy cost per position fix for each of the three ST87M01 positioning modalities, and how does this vary across outdoor, urban, and transitional scenarios?
• How significantly does A-GNSS reduce acquisition energy and time-to-first-fix relative to a cold start, and under what conditions is this benefit practically relevant?
• What is the accuracy of NB-IoT Cell-ID and Wi-Fi positioning under realistic conditions, and for which tracking use cases is each modality sufficient?
• How accurately does an analytical energy model predict total system consumption across different duty-cycle configurations?
• Which operating state dominates the energy budget, and what are the most impactful parameters for extending battery lifetime?

Expected Contribution

• A comprehensive characterisation dataset for all three positioning modalities of the ST87M01, representing a contribution to the sparse independent literature on this module
• A validated analytical energy model covering all relevant operating states, with measured parameterisation on the ST evaluation platform
• Concrete design guidelines for duty-cycle configuration and technology selection in low-power tracking systems based on the ST87M01
• A directly usable foundation for the accompanying master's thesis, which extends this work with a custom hardware design and an adaptive technology selection strategy

Contact: Markus Cremer ([email protected])