Biomechanical cost quantification for wheelchair accessible cities – CapaCITIES

The project relied on a biomechanical evaluation of MWC users performing various scenarios of controlled displacements on a dedicated MWC simulator. This has required the development of tools for biomechanical evaluation, the development of a haptic and immersive MWC simulator, and the ability to aggregate several biomechanical parameters with different units in a single index.
In terms of biomechanical evaluation, the objectives were: 1) to implement a musculoskeletal model including a close loop definition of the shoulder kinematic chain, and to associate muscle paths; 2) to provide a scaling method to obtain subject-specific models without resorting to medical imaging; and 3) to implement a computationally efficient method for the assessment of the biomechanical parameters that enable to sort the environmental obstacles in terms of physical requirements. Besides, to collect the biomechanical data under realistic propulsion conditions and in a controlled manner, we have chosen to reproduce MWC propulsion on a simulator. However, the realism of the various current simulators was limited either by i) low immersion, ii) the absence of haptic feedback, or iii) the lack of sensors needed to reproduce the mechanical behaviour of the MWC or to quantify the biomechanical parameters. Faced with these limitations, our aim was to develop a dynamic, haptic and immersive simulator based on a Stewart platform and a roller ergometer. Another major original feature of this simulator was the possibility of changing the MWC settings by remote control while the user was still seated on the simulator, since the MWC configuration has a major impact on mobility. The sensors on board the ergometer enabled useful data to be measured for both biomechanical analysis and simulator control, including haptic, visual and vestibular feedback. Finally, in terms of biomechanical cost, several indices had already been proposed in the literature, but no study had sought to quantify MWC accessibility on a continuous scale (other than 0 vs 1), by associating a cost attributed to different environmental situations. Furthermore, depending on the situation, the relevant biomechanical parameters for classifying the difficulties of the situations may be different (work vs. joint stress, for example). Consequently, it was necessary to be able to associate costs calculated differently, i.e. on the basis of different biomechanical parameters, depending on the situation, and ideally to provide a single index integrating the different aspects of the physical demands of the environmental situations to be crossed.

As part of the CapaCITIES project, we developed a manual wheelchair (MWC) simulator, which enhanced the PSCHITT-PMR platform at LAMIH to study MWC locomotion in a controlled and secure virtual environment under various conditions. This facilitates the evaluation and quantification of user performance. The simulator consists of six screens providing a 150° immersive field of view, a motorized roller ergometer accommodating an instrumented MWC, and a six-degree-of-freedom hexapod platform. It incorporates visual, haptic (to simulate resistances to motion), and dynamic (by reproducing acceleration sensations via the hexapod) feedbacks. A haptic and dynamic interface, based on a model of interactions between the user, the MWC, and the ground, was developed to reproduce MWC behavior in real-time. A robust haptic controller ensures precise tracking of reference velocities while generating realistic haptic forces.

A biomechanical model of the upper limb, based on various existing models from the literature and incorporating closed-loop constraints, was integrated into the CusToM library to analyze MWC propulsion. Implementing geometric closures at the shoulder complex led to the development of specific resolution methods for the kinematic and kinetic analysis of MWC propulsion movements. We also adapted model customization tools (geometric scaling) to these new constraints and worked on computation time optimization, as increased model complexity directly impacts processing time.

Experimental campaigns were conducted using the simulator, where participants equipped with reflective markers performed various locomotion scenarios (including slopes, cambers, and turns). These experiments allowed for the collection of kinematic and kinetic data, including handrim torques and joint kinematics. A preliminary assessment of participants' physical capacities was also conducted.

The analysis of data under slope conditions revealed linear correlations between biomechanical parameters such as push time, push angle, handrim torques, and average velocity with slope inclination. However, correlations with joint kinematics were weak, except for certain degrees of freedom. These results suggest the possibility of inferring specific biomechanical parameters from obstacle characteristics. Furthermore, preliminary work based on minimizing energy expenditure due to rolling resistance has demonstrated the relevance of path planning optimization approaches.

At the end of the project, several improvements can be considered across the different parts of the project.

Regarding the biomechanical evaluation, despite the improvements in computation time for musculoskeletal analysis, challenges remain for the most complex models, particularly in terms of solution stability, problem convexity, and optimization complexity. The integration of neural networks trained to resolve these loops could significantly reduce computation time. As for the biomechanical model itself and its customization methods, further research is needed, as evidenced by the growing number of scientific publications on these issues. The validation of this customized model at different levels (kinematic, inertial, and muscular) remains a challenge for the scientific community.

Regarding the manual wheelchair propulsion simulator, various enhancements are also possible. Indeed, during the experiments, some expert users reported a lack of realism in the simulator due to the absence of trunk and upper limb movements affecting the wheelchair’s kinematics. Addressing this will require additional instrumentation and the development of more precise behavioral models. Furthermore, while it is possible to modify the wheelchair’s physical settings, integrating the impact of these characteristics into the control model remains challenging, particularly because many parameters are identified at different levels. However, considering the interaction between environmental effects and wheelchair characteristics is crucial. Rewriting the model to allow for rapid and reliable modifications represents an important future direction.

In terms of route optimization, three key perspectives emerge. The first is to investigate how to infer biomechanical parameters from the physical characteristics of obstacles. While the results for slopes are promising, they need to be extended to other locomotion tasks. The second perspective involves mapping the environment based on biomechanical cost, which will require efficiently capturing the physical characteristics of obstacles across an entire neighborhood or city. Finally, the third perspective focuses on comparing the biomechanical cost of obstacles with users' physical and technical abilities. Considering the effects of user fatigue on their physical and technical capabilities is also crucial to ensure that the proposed solutions remain relevant regardless of the route length.

Publications :
• Livet et al. A penalty method for constrained Multibody kinematics optimisation using a Levenberg-Marquardt algorithm. CMBBE, 2022, pp.1-27. (hal-03697267)
• Rouvier et al. Manual wheelchair biomechanics while overcoming various environmental barriers: A systematic review. PLoS ONE, 2022, 17 (6), pp.1-21. (hal-03703872)
• Livet et al. An automatic and simplified approach to muscle path modelling. Journal of Biomechanical Engineering, 2022, 144 (1), pp.1-9. (hal-03279707)

Communication:
• Livet et al. Expected scapula orientation error regarding scapula-locator uncertainty while studying wheelchair locomotion. Congrès de la Société de Biomécanique, Saint-Etienne, France, 2021. (hal-03276057)
• Louessard et al. A preliminary investigation of handrim kinetics in various environmental situations crossed in manual wheelchair. Congrès de la Société de Biomécanique, Saint-Etienne, France, 2021 (hal-03331000)
• Rouvier et al.. Comparison of scapula soft tissue artefact compensation methods during manual wheelchair locomotion. ESMAC congress, Dublin, Ireland, 2022. (hal-03794174)
• Bentaleb et al. Numerical Simulator for Manual Wheelchair Propulsion based on a MPC Approach. 4th IFAC Workshop on Cyber-Physical and Human Systems CPHS 2022, Dec 2022, Houston (Texas), United States. pp.154-159 (hal-04481301)
• Ait-Ghezala et al., Haptic Interface Design for a Novel Wheelchair Simulator using Linear Time-Varying MPC Framework. 2023 IEEE/ASME International Conference on Advanced Intelligent Mechatronics (AIM), Jun 2023, Seattle (USA), Washington, United States. pp.172-178, (hal-04481276)
• Ait Ghezala et al. Haptic Wheelchair Simulator: Design with Experimental Validation. Driving Simulation Conference Europe 2023 VR, Sep 2023, Antibes (06), France. (hal-04631738)
• Livet et al. Open vs closed articular architecture of the forearm for an analysis of muscle recruitment during throwing motions. ISB 2021 - XXVIII Congress of the International Society of Biomechanics, Jul 2021, Stockholm, Sweden. pp.1. (hal-03241267)
• Ait Ghezala et al. Direct Model-Reference Adaptive Control for Wheelchair Simulator Control via a Haptic Interface. 15th IFAC Symposium on Analysis, Design and Evaluation of Human Machine Systems HMS 2022, Sep 2022, San Jose (CA), United States. pp.49-54, (hal-04481305)
• Demestre et al. Analyse biomécanique de la propulsion en fauteuil roulant manuel lors de la locomotion sur des pentes dans un environnement simulé. Conférence Handicap 2024, Jun 2024, Paris, France. (hal-04550571)

Thèses de dotorat
• Claire Livet. Contributions algorithmiques à l'analyse musculo-squelettique : modèles et méthodes. (tel-03765953)
• Théo Rouvier. Approche biomécanique pour quantifier l’accessibilité en fauteuil roulant manuel. (tel-04174904)
• Amel Ait Ghezala-Sadoudi. Assistance haptique pour la propulsion en fauteuil roulant manuel (FRM) basée sur une évaluation des capacités biomécaniques de l’usager. (tel-04942682)

Locomotion with a manual wheelchair (MWC) submits the upper-limbs of the manual wheelchair users (MWU) to an important stress, which varies according to the environment. To assist MWU in selecting the paths that preserve their upper limbs, a cost reflecting the physical demand of the successive situations along the possible paths must be attributed. In the current state of knowledge and accessibility standards, an obstacle has no graduation and can only be marked as crossable or not, which cannot reflect, neither the heterogeneity of the situations, nor the link between their accessibility and the physical and technical abilities of the MWU. To go beyond these limitations, this project aims at defining biomechanical costs that can be attributed to the environmental situations, and that could be implemented in future optimal path selection algorithms. This will make it possible to provide MWU with individualized paths taking into account their individual capacities. To do so, a musculoskeletal model will be developed to quantify various biomechanical quantities that will serve as input data for the definition of the biomechanical costs. These costs will be computed for various situations, reproduced in a realistic MWC locomotion simulator developed in the framework of this proposal. Such a project will provide original and useful data for accessibility evaluation, planning of urban development services and assistance adaptation. It will also be the basis for further work on MWU evaluation and paths characterization to provide personalized cost-optimal paths.