Search

International audience; Obstacle avoidance systems for autonomous vehicles combine multiple sensing technologies (i.e. LiDAR, Radar, Ultrasound and Visual) to detect different types of obstacles across the full range of lighting and weather conditions. Sensor data are fused with vehicle orientation (obtained for instance from an Inertial Measurement Unit and/or compass) and navigation subsystems. Power hungry, they require powerful computational capability, which limits their use to high-end vehicles and robots. 2 INSPEX ambition The H2020 INSPEX project plans to make obstacle detection capabilities available as a personal portable multi-sensors, miniaturised, low power device. This device will detect, locate and warn of obstacles under different environmental conditions, in indoor/outdoor environments, with static and mobile obstacles. Potential applications range from safer human navigation in reduced visibility conditions (e.g. for first responders and fire brigades), small robot/drone obstacle avoidance systems to navigation for the visually and mobility impaired people. As primary demonstrator (Fig.1), we will plug the INSPEX device on a white cane (see Fig. 1) for Visually Impaired and Blind (VIB) people to detect obstacle over the whole person height, provide audio feedback about harmful obstacles, improve their mobility confidence and reduce injuries, especially at waist and head levels [1]. The device will offer a "safety cocoon" to its user.

International audience; Obstacle detection capability is of great interest in application domains such as navigation of Visually Impaired and Blind people. To perform free-space assessment over the whole person height, range sensors are usually placed on a white cane and their measurements are analyzed to provide feedback about potential harmful obstacles. Thanks to technology progress both at hardware and software levels, integration of such obstacle detection capabilities in a white cane seems conceivable. It is well admitted that ultra-sonic range sensors have limited sensing range (typically < 3 m) and difficulties of operating on highly reflective surfaces [1]. Laser-based solutions can be highly sensitive to ambient natural light and identification of transparent or mirror-like surfaces is difficult. RF Radar range sensor performance is affected by the electromagnetic backscattering characteristics of the obstacle (Radar Cross Section). The electromagnetic response is in general very different from the mechanical response to ultrasound waves or optical response to LiDAR. [2] shows that Ultra-Wide-Band (UWB) radar can be used effectively to detect obstacles under rain, snow, fog and smoke, thus being complementary to LiDAR. Therefore, to overcome limitations of each range sensor technology, Ultra-Sound, UWB RF Radar and LiDAR will be co-integrated in the obstacle detection system the H2020 INSPEX project is targeting (Fig. 1) [3]. This work details the characterization and future enhancements of the UWB RF Radar developed in the course of the project. The complete system should not exceed 200gr in weight and 100cm3 in volume. Ten hours of lifetime in continuous use are expected with an initial target for power consumption smaller than 500mW. The obstacles could move at an unpredictable speed and direction and the system should detect reliably objects moving at a speed of ~1.4 m/s worst case (relative speed between the user and the objects) and be able to detect obstacles up to a distance of 4m. After a static characterization phase of the UWB Radar, measurements were performed in mobility conditions. The Radar is placed on a cart moving linearly towards the obstacle with a colliding trajectory. The relative radial speed is about-0.15m/s. The right drawing in Fig 1 shows the Radar response over time with three snapshots in the speed domain. The top left curve clearly shows the “approaching” obstacle wave-front while the three other curves showhow the speed can be exploited in colliding trajectories detection.

International audience

International audience

Obstacle avoidance systems for autonomous vehicles combine multiple sensing technologies (i.e. LiDAR, Radar, Ultrasound and Visual) to detect different types of obstacles across the full range of lighting and weather conditions. Sensor data are fused with vehicle orientation (obtained for instance from an Inertial Measurement Unit and/or compass) and navigation subsystems. Power hungry, they require powerful computational capability, which limits their use to high-end vehicles and robots.