Ultralight Autonomous Excavator

An Ultralight Autonomous Excavator (UAE) is proposed that is capable of performing common earthwork tasks with minimal supervision. The UAE is an 80kg excavator which relies on the dynamics of a demolition hammer to break up terrain before scooping it away with a bucket. The UAE features an integrated excavation arm, demolition hammer, backfill blade, and a battery bank that doubles as a counter weight. The addition of a demolition hammer trades the prerequisite mass required for excavation with a well-known and predictable mechanical system. A demonstration-based teaching system, The Macro Learning Platform, is concurrently proposed to enable novel earthwork tasks with lean training data. The teaching system was inspired by experienced machine operators feeling machine responses to anticipate different excavation conditions. The Macro Learning platform learns the UAE mechanical responses as driven by a series of user inputs and allows the responses to be extrapolated over many dimensions. The platform provides a user the opportunity to define more fundamental and intuitive quantities in the excavation process, such as hammer traction and downforce, rather than just arm motion. The UAE software stack is built on the ROS 2.0 framework. The code base is modular and relies only on the ROS communication interface to share data. The hardware stack is assembled from COTS components (except some brackets and the bucket). The COTS hardware provides known durability, inexpensive and fast repairs, and maximal operational availability. Phase 1 work will drive towards the demonstration of an MVP of cut and fill operations and the repair of a simulated blast crater.