| Summary: | Wheelchairs are vital tools for people with disability which allow them to regain the ability to move again. A Shared Control Robotic Wheelchair further reduces the effort needed by the users bringing them a more comfortable experience. One key aspect of a Shared Control Robotic Wheelchair is the usage of a Navigation Among Crowd Algorithm (NACA) also known as a Collision Avoidance Algorithm (CAA). With the implementation of NACA to the Shared Control Robotic Wheelchair, the user has control authority over where they want to move to and avoid collision with pedestrians and obstacles. This brings great convenience to users who have disability. Some of the common NACA are Dynamic Windows Approach (DWA) or Timed-Elastic Band (TEB) etc. However, with new NACA being introduced and improvements being made as technology advances, it is difficult to determine which algorithm is the most suitable for implementation to a Shared Control Robotics Wheelchair. There are evaluation frameworks evaluating the performance of NACA. However, the existing evaluation framework mainly focuses on physical environment instead of crowd which does not truly evaluate NACA in shared control. Therefore, a framework is needed for evaluation of NACA in shared control setting. In this report, three different types of NACA were chosen for evaluation of their benefits and limitations: Reactive Driving Support algorithm (RDS), Shared Dynamic Windows Approach (Shared-DWA) Static Model and Constant Velocity Model. RDS was integrated into the existing wheelchair shared control system by RRIS. Evaluation frameworks were then developed with different scenarios focusing on crowd movement and consist of different metrics that are assigned into 4 categories for evaluation. Experiments were carried out in both Simulation and Real-Life Demo with Data collection. Lastly, with evaluation tool, the benefit and limitations were evaluated. Overall, SDWA-S performs better overall considering safety and comfort of wheelchair user even though RDS came close to it. The results obtained debunked the hypothesis that NACA meant for dynamic obstacle will perform better than NACA that treats moving obstacles as static in a crowd. The report also concluded that pedestrian should be considered static since SDWA-S method treats dynamic obstacles as static and it performed the best.
|
|---|