Project Overview
The Advanced 9-Axis Inertial Measurement Unit (IMU) Project was a multidisciplinary R&D initiative focused on developing high-performance inertial sensing, motion estimation, and sensor fusion technologies.
The primary objective was to develop both standalone and network-based 9-axis IMU architectures capable of accurately estimating orientation and motion under dynamic operating conditions.
The project combined hardware development, embedded software engineering, advanced signal processing, sensor fusion algorithms, wireless-independent wired sensor networking, and experimental validation on a quadruped robotic platform.
A key aspect of the project was the transition from conventional single-sensor inertial measurement systems to a distributed, network-based multi-IMU architecture. This approach enabled the integration of measurements from multiple sensor nodes to improve motion estimation accuracy and robustness.
The project commenced in November 2023 and progressed through hardware development, algorithm design, software integration, robotic system implementation, experimental testing, and technical documentation activities.
Project Development Timeline
-
November 2023 – January 2024
1. IMU Hardware Research and Sensor Architecture Selection
The initial phase focused on evaluating alternative 9-axis inertial measurement sensor platforms and identifying suitable hardware architectures for the intended applications. Three major sensor platforms were investigated:
- Arduino Nano 33 BLE Sense Rev2, featuring a 9-axis inertial sensing architecture.
- Bosch BNO055 9-axis inertial measurement sensor, including evaluation of the Bosch BSX 3.0 Sensor Fusion Software library.
- TDK InvenSense MPU-9250 / MPU-9255 9-axis inertial measurement sensors.
The activities included hardware architecture analysis, schematic evaluation, sensor characteristics assessment, and investigation of suitable microcontroller architectures for processing inertial sensor measurements.
MATLAB-based PC interface libraries and preliminary application software were also investigated and developed. In parallel, Bill of Materials (BOM) analyses were performed to evaluate the cost implications of alternative hardware configurations.
-
February – March 2024
2. Modular Sensor Processing and Orientation Estimation
The second phase focused on developing and evaluating modular approaches for inertial sensor processing and orientation estimation.
Initially, a conventional orientation estimation method was implemented by combining angular velocity integration from the gyroscope with acceleration measurements from the accelerometer for inclination correction. The resulting orientation estimates were represented using quaternion notation. This implementation established a reference benchmark against which the subsequent modular processing architecture could be evaluated.
In March 2024, accelerometer and gyroscope processing algorithms were separated into independent modules. Unlike the conventional approach, the modular architecture did not include a feedback loop from the inclination correction stage to the IMU integration process. The resulting performance was compared against the previously developed conventional method.
-
April – May 2024
3. Advanced IMU Sensor Fusion and Signal Processing
A major challenge in inertial measurement systems is the accumulation of orientation estimation errors caused by gyroscope bias, measurement noise, and integration drift. To address these challenges, advanced signal processing and sensor fusion methods were developed.
In April 2024, measured acceleration signals were decomposed into their individual components, and second-order Butterworth low-pass filters were applied to each component. This approach enabled effective averaging of accelerometer measurements while allowing short-term acceleration and deceleration effects to compensate for each other.
In May 2024, each accelerometer measurement was represented as a three-dimensional vector and compared against a vertical reference vector. This method enabled the separation of gravitational acceleration from acceleration components associated with changes in motion velocity, improving the reliability of orientation estimation.
-
June – September 2024
4. Gyroscope Bias Estimation and Adaptive Magnetic Disturbance Rejection
This phase addressed two critical sources of error in inertial orientation estimation: gyroscope bias and magnetic disturbances.
Gyroscope Bias Estimation
In June 2024, algorithms were investigated to identify periods during which the IMU remained stationary, enabling the estimation of sensor bias under stationary conditions. Kalman filtering approaches based on Gaussian distributions were evaluated, and the mathematical model of the filter was established.
In July 2024, filter parameters were optimized to ensure that bias estimates obtained during stationary periods were assigned higher confidence than estimates obtained during motion. The estimated bias values were subsequently compensated for in gyroscope measurements, improving the accuracy of orientation estimation.
Adaptive Magnetic Disturbance Rejection
In August 2024, an adaptive filtering algorithm was developed to minimize the effects of temporary magnetic disturbances on inertial orientation estimation. Two alternative magnetic disturbance detection approaches were investigated:
- Monitoring deviations in the magnitude of the local magnetic field.
- Evaluating deviations in magnetic inclination using user-defined or automatically determined reference values.
In September 2024, the disturbance rejection algorithm was further developed. Magnetometer measurements were considered reliable when they remained sufficiently close to the reference values for at least 0.5 seconds. Under detected magnetic disturbance conditions, magnetometer updates could be disabled for periods of up to 60 seconds.
This approach improved the robustness of orientation estimation in environments affected by temporary magnetic interference.
-
October – December 2024
5. Software Integration and 9-Axis Sensor Fusion
The next phase focused on integrating the previously developed algorithms into a unified sensor processing architecture.
In October 2024, the first integrated implementation consisted of a 6-axis IMU configuration combining three-axis accelerometer and three-axis gyroscope measurements. Simulation studies were initiated to evaluate the integrated motion estimation algorithms.
Although the 6-axis configuration provided accurate results over short measurement periods, accumulated integration errors could cause drift during longer operating periods. To address this limitation, a three-axis magnetometer was integrated into the sensor architecture in November 2024, establishing the complete 9-axis IMU configuration.
In December 2024, MATLAB/Simulink Embedded Coder tools were utilized to convert the developed algorithms into embedded software suitable for deployment on microcontroller platforms.
-
January – March 2025
6. Embedded Implementation and Experimental Validation
The sixth phase focused on implementing and validating the developed 9-axis IMU algorithms on physical hardware platforms.
In January 2025, test and verification procedures were defined for the Arduino Nano 33 BLE hardware platform. Dedicated motion test prototypes were also developed to support systematic performance evaluation.
In February 2025, embedded software development and implementation activities were extended to the Bosch BNO055 9-axis sensor platform. Interface development between the implemented algorithms and the Bosch BSX 3.0 software library was also performed.
In March 2025, experimental motion tests were conducted at the Robotics Laboratory of Istanbul Technical University (ITU), Faculty of Electrical and Electronics Engineering. The laboratory's motion testing infrastructure was utilized to evaluate the algorithms under controlled movement conditions.
Additional arrangements were made to generate magnetic disturbances in the laboratory environment. The developed algorithms were evaluated under these conditions, demonstrating their ability to produce reliable orientation estimation results in the presence of magnetic interference.
-
April – June 2025
7. Development of a Network-Based 9-Axis IMU Architecture
A major architectural evolution took place in April 2025 with the transition from standalone 9-axis IMU units to a network-based distributed measurement system. A conceptual architecture was established, and the fundamental hardware requirements for communication and integration between multiple IMU nodes were defined.
This development introduced a distributed inertial measurement architecture in which multiple 9-axis IMU nodes could operate as part of a unified measurement system.
In May 2025, I2C serial communication was initially implemented and tested for communication between sensor nodes. However, performance limitations were observed in environments where external disturbances were significant. Consequently, alternative serial communication architectures were investigated.
In June 2025, RS-485 was selected as the inter-node communication protocol to improve communication robustness. A CRC-based data integrity mechanism was also incorporated to support reliable data transmission and error detection.
With these developments, the hardware design activities for the network-based 9-axis IMU system were completed.
-
July – September 2025
8. Multi-Sensor Fusion, Extended Kalman Filtering and Proprioceptive Odometry
The project subsequently focused on developing the software architecture required to process and combine measurements from multiple networked IMU nodes.
In July 2025, transformation and connection matrices were developed to define the spatial relationships between individual 9-axis IMU nodes. An Extended Kalman Filter (EKF) architecture was established to combine measurements from the different sensor nodes within a centralized processing unit. Generic motion and sensor models were initially defined for the filter architecture.
In August 2025, the application was extended to four-legged walking robots (Quadruped Robots). The objective was to estimate the position and orientation of the robot's main body using onboard sensing information, including networked IMUs and joint encoders. This approach, known as Proprioceptive Odometry (PO), utilized measurements from the distributed 9-axis IMU network and integrated them within a centralized EKF framework.
In September 2025, the network-based Proprioceptive Odometry architecture was completed. An XML-based fundamental user interface was also developed to facilitate the integration of generic motion and sensor information into the system. The interface architecture was designed to support further development and expansion.
-
October – December 2025
9. Network-Based IMU System Testing on a Quadruped Robot
The final experimental development phase focused on evaluating the performance advantages of the network-based 9-axis IMU architecture compared with a conventional standalone IMU configuration.
A quadruped robotic platform at the Robotics Laboratory of Istanbul Technical University was selected as the experimental test platform. An OptiTrack optical motion capture system was used to obtain highly accurate reference position and orientation measurements. These measurements served as Ground Truth data for evaluating the performance of the developed network-based IMU system.
Static and Controlled Orientation Tests
In November 2025, the robot's body orientation was changed along three rotational axes: roll, pitch, and yaw. The measurements obtained from the network-based IMU system were compared against the OptiTrack reference measurements. The results demonstrated improved performance compared with the configuration using a single IMU mounted on the robot's main body.
Dynamic Walking Tests
In December 2025, the validation activities were extended to dynamic walking conditions. The experiments were conducted within a limited 5 m × 5 m test area due to the operational constraints of the optical motion capture system.
The network-based IMU architecture was evaluated under complex robot movement conditions, and successful motion estimation results were obtained. These experiments provided further evidence of the benefits of distributed inertial sensing and multi-sensor fusion for robotic motion estimation.
-
January – May 2026
10. Technical Data Package (TDP) and Project Documentation
The extended project phase focused on consolidating the engineering outputs and preparing comprehensive Technical Data Package (TDP) documentation for both standalone and network-based IMU systems.
Standalone IMU Technical Documentation · January – March 2026
The documentation process for the standalone 9-axis IMU system was initiated in January 2026. Hardware design documents, software modules, and test and verification reports were classified and organized according to the TDP structure.
During February 2026, reports corresponding to individual work packages, algorithm descriptions, and embedded software outputs were integrated into the documentation package. Test scenarios and verification results were also converted into standardized reporting formats.
In March 2026, the final documentation review and consolidation activities were completed, including traceability improvements, document version control, and file coding structures. The standalone IMU Technical Data Package was thereby consolidated into a comprehensive documentation structure.
Network-Based IMU Technical Documentation · April – May 2026
In April 2026, the TDP preparation process for the network-based 9-axis IMU system commenced. Experimental data obtained from the Robotics Laboratory tests were analyzed and documented. Technical documentation covering the network communication architecture and EKF-based sensor fusion algorithms was also organized.
In May 2026, the network-based IMU documentation reached its finalization stage. Detailed reports covering static and dynamic quadruped robot tests were completed, including comparisons against Ground Truth measurements.
The technical documentation was converted into its final format in preparation for project closure and delivery.
Engineering Capabilities and Project Achievements
The IMU Project demonstrates a comprehensive engineering approach combining advanced inertial sensing, embedded systems, signal processing, multi-sensor fusion, and robotic motion estimation. The major engineering capabilities developed and applied throughout the project include:
- 9-axis inertial measurement system architecture
- Accelerometer, gyroscope, and magnetometer integration
- Sensor characterization and hardware evaluation
- Embedded software development using MATLAB/Simulink Embedded Coder
- Quaternion-based orientation estimation
- Advanced digital signal processing and Butterworth filtering
- Kalman filtering and gyroscope bias estimation
- Adaptive magnetic disturbance detection and rejection
- Distributed multi-IMU network architecture
- RS-485 communication and CRC-based data integrity
- Extended Kalman Filter (EKF) implementation
- Proprioceptive Odometry for quadruped robotic platforms
- Multi-sensor spatial calibration and transformation matrices
- Ground Truth-based experimental validation using optical motion capture
- Technical Data Package preparation and engineering documentation
Conclusion
The Advanced 9-Axis IMU Project represents a significant R&D effort in the development of intelligent inertial measurement and multi-sensor motion estimation technologies.
Beginning with the evaluation and implementation of standalone 9-axis IMU platforms in November 2023, the project evolved into a distributed sensor network architecture capable of combining measurements from multiple inertial sensing nodes.
Through the development of advanced filtering algorithms, Extended Kalman Filter-based sensor fusion, and Proprioceptive Odometry, the system was extended to robotic motion estimation applications and experimentally evaluated on a quadruped robotic platform.
The use of optical motion capture measurements as Ground Truth references enabled comparative performance evaluation under both static and dynamic operating conditions.
The project subsequently progressed through comprehensive technical documentation and Technical Data Package preparation, covering both standalone and network-based IMU systems.
This project demonstrates ARBON's engineering capabilities in advanced sensor technologies, embedded systems, distributed measurement architectures, sensor fusion algorithms, and robotic motion estimation — bringing together hardware, software, and system-level validation within a unified R&D development process.