Main computer (behavior and high level control): Mini ITX board (GIGABYTE GA-H87N-Wifi) with Intel Core i7 4x3.0 GHz, 8 GB RAM, Ethernet and CAN communications. Vision computer (object detection, visual odometry): Intel NUC i7. Leg controller (low level motor control): UCOM Universal Controller Module, custom with DSP and FPGA providing real-time control, quadrature encoder readout, current measurements, initialization routines, synchronization, and communication. Additional controller: Arduino for light control and Raspberry Pi for 3D laser scanner.
Ubuntu 14.04 with ROS Indigo for high level control and MCA2 for low level control. High level control is done in the PLEXNAV framework, providing OMPL-based Planning, OpenCV, and PCL based object and environment recognition, SMACH based mission control, 3D SLAM with custom algorithms, custom motion executors and many many more packages both custom written and open source. Low level control is implemented in the custom MCA2 (Modular Controller Architecture) Framework using a behavior based approach. Fast behaviors govern the leg moments (swing, stance, hit obstacle reflex) while higher level behaviors keep the overall system stable. The motor controllers run cascaded PID position controllers.
Two 24-V, 8000-mAh lithium-polymer battery packs in parallel, 2 hours of operation. Ability to hot swap battery packs during operation. External supply by a laboratory power supply at 24-V (5-20 A, ~10A main power) can be used on the fly.