Dancing robots: Experiments in domotics with balancing robots and high-performance sensor fusion

Details

Supervisors

Van Roy, Peter

Faculty

Ecole polytechnique de Louvain

Degree label

Master [120] en sciences informatiques, à finalité spécialisée

Abstract

enDomestic automation has been quietly accumulating sensor-rich devices that share indoor space with the people living there, and mobile robotics is converging on the same set of problems. Two-wheeled self-balancing robots sit in the middle of that space: compact, mechanically unstable, and constrained in a way that forces a particular kind of design discipline. This thesis builds on top of an inherited self-balancing chassis, brought to a working balancing state on a GRiSP-2 board running Erlang/OTP by a previous master thesis, and asks what to do with the robot once it is standing. The longer-term picture is a small robot that follows a second, taller robot around a home without GPS, without external cameras, and without beacons of its own. To get there, three intertwined problems are taken up: a path-following layer that traces a smooth curve through received waypoints, a wireless waypoint transmission protocol that shares the inherited single-byte radio link without breaking the existing manual control, and an onboard pose estimator that works only from wheel speeds. Four pieces of work are reported, none of which modifies the inherited balance loop. A spline+tracker stack provides smooth waypoint following, with a small sub-state machine for new waypoint arrivals. An over-the-air protocol carries waypoints from a sender. An empirical study compares three onboard localization configurations for yaw (wheel odometry alone, yaw based on the integrated gyro, and yaw based on tilt-compensated magnetometer) on the same chassis and the same arena; and a lifecycle finite-state machine that manages the overall trajectory state and with a colour-coded LED scheme makes the system state legible to external observers.