The objective of the project is to find an intelligent way to control a powered wheelchair, in order to prevent its users from pit falling, for example on a downward staircase or hole, and from tipping, due to a bump, sidewalk or wall.

To accomplish this, the first step would be to identify the environment in front of the chair and label it as a safe or dangerous situation. For this task, 2 different types of depth sensors were characterised, TeraRanger Evo64 px and StereoLabs ZED2 Depth Camera.

For the Terabee Evo 64 px the experiments and their results were not accurate nor satisfatory for the level of precision needed in a wheelchair. To verify it's reading fluctuation, a further test was done with a comparison of the readigns of a sensor in a white and black surface. The results of this experiment can be visualized in the following histograms.

Besides all these problems, the frame rate varies according to the field in front of it which is a big problem since it is needed that the sensor is able to respond to every situation equally.

Based on these, the best choice is the ZED2 Camera due to it's high performance, accuracy and stability.

In conclusion, after comparing both sensors and evaluating both hypotheses, the ZED2 Depth Camera clearly showed to be more promising due to it’s high resolution, wide field of view and stable readings throughout time and independent of the situation.

Moreover, the work developed with the ZED2 already allows to identify multiple dangerous situations, as exceed inclination, bump detection and pitfall detection (important to mention that these tests could only be tested inside the house, in a limited space, due to the Covid19 pandemic).

As future work, the plan is to conduct the test in an outside environment with real life situations, to fine tune the algorithm parameters. After that, the last step would be to attach the sensor to the wheelchair, in order to control the wheelchair speed according to the classified situation.