A low-power, high-dynamic range sensor system for long-term autonomy, navigation, and mapping on the lunar surface

Tangram Vision will develop a low-power sensor system for a Lunar Terrain Vehicle (LTV) that utilizes a wide-FOV stereo pair of event cameras (a.k.a. neuromorphic camera or dynamic vision sensor), along with an IMU, to enable quick localization and confident mapping in environments with extreme lighting conditions. This combination of novel sensor modalities will enable the creation of an event-informed dense depth map of the environment that is consistent over time and lighting conditions, due to an event camera’s ability to identify occlusions and objects in motion relative to the world coordinate frame. Funding will be used to create a hardware prototype of the proposed sensor system. The remaining capital will be allocated to the technical experts for their investigations and contributions to the project. At the end of this phase, Tangram Vision will have demonstrated that a stable depth synthesis pipeline can be created from two stereo event cameras. If a stable pipeline has not been achieved by the end of this phase, Tangram Vision will report how such a pipeline could be achieved given different resources or approaches. Target markets for such a system extend well beyond lunar exploration. A sensor suite like the one proposed has applications in robotics, autonomy, drones, and any system moving at a high speed through challenging lighting conditions. We see this work as the first step towards fully commercialized event sensing, which currently does not have a presence in the sensor market.