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

This work proposes a stereo event camera rig for robust mapping and navigation on the lunar surface. Initial testing in Phase I validated our system’s ability to generate accurate depth maps from event data alone. New calibration techniques specific to event cameras were developed, adapting traditional computer vision approaches to handle asynchronous event data streams. In sum, this configuration shows great promise for use on autonomous and teleoperated systems operating on the lunar surface, where harsh lighting conditions are common and information is sparse. Phase II development will introduce a self-calibration regime for all sensors on the rig, allowing the system to produce accurate data over longer periods of time. Tangram Vision will also test the abilities of the proposed sensor system extensively on autonomous platforms in harsh terrestrial conditions as a proxy for the lunar surface. Tangram Vision intends to use the funding largely to develop and refine these novel self-calibration routines, depth synthesis algorithms, and odometry algorithms for the purpose of long-term autonomy. By the end of Phase II, our sensor system will have been tested on an autonomous platform in a lunar-like environment (the deserts of California). The majority of the grant will be devoted to algorithmic development, alongside integration testing with deployed platforms. Target markets for this innovation include any terrestrial or extraterrestrial system that is expected to behave autonomously for long durations and/or in changing environments. This includes space exploration systems (lunar rovers, autonomous landing systems, and other forms of space robotics), defense and military robotics (autonomous ground vehicles, systems operating at night, and high-speed platforms), and commercial autonomous systems (industrial robotics, mining, construction, humanoids, last-mile delivery, mobility, etc.).