The REXASI-PRO 3D Driving Assistant (3DDA) is a software module that safeguards driving commands for electrically powered wheelchairs. Driving commands can either be issued by a human driver using a standardized joystick or generated by the navigation approach to be developed in REXASI-PRO, which is based on neural networks. 3DDA integrates perceptions of two physical laser scanners as well as virtual laser scans derived from depth imaging cameras into an occupancy map. This map is continuously tested against a safety area for possible collisions. The safety area is dynamic, which means that its shape adapts to the current speed, the driving commands, and the physical properties (e.g. maximum acceleration and deceleration) of the wheelchair. If the 3DDA detects a possible collision, the current driving command is modified in such a way that the obstacle is avoided, or a complete stop is initiated.
To accommodate new use cases, new wheelchairs are needed. Currently we are in the process of equipping electrically powered wheelchairs with hardware components necessary for safe software supported navigation. To achieve this goal, the wheelchairs are equipped with a front and rear laser scanner, incremental encoders for both motorized wheels, and two depth imaging cameras. Additionally multiple small microcontrollers are added to interface with the wheelchair’s CAN-Bus and the rotary encoders respectively. For the purpose of better maintainability, and to enable easier reproducibility of the results, all the hardware used is off-the-shelf commodity hardware like Arduinos and prebuilt CAN-Transceivers. This allows easy replacement and programming without the need for special hardware. Another benefit of prebuilt hardware is a fast turnover during upgrades and tests because no special PCBs are needed.