Mission Control coordinates Isaac AMR Cloud Services to send robots on missions. When a mission is submitted to the Mission Control API, it uses map data provided by DeepMap, a route graph from the Waypoint Graph Generator, and multi-robot route optimization results from cuOpt to assemble a behavior tree of tasks. Then, Mission Dispatch executes the behavior tree by sending commands to the Mission Client on the robot following the VDA5050 standard.
This version of Mission Control only supports one robot.
DeepMap serves occupancy grid maps and semantic maps to Isaac AMR Cloud Services and the robots. An occupancy grid map discretizes the environment into sections, and each section is assigned a value representing the probability that it is occupied by an obstacle. A semantic map describes the semantic properties of different sections of the map with polygon labels.
A waypoint graph is a collection of waypoint nodes–which are points in the free space of the map–edges connecting the nodes, and finite costs associated with the edges. The waypoint graph generator can construct waypoint graphs on occupancy grid and semantic maps.
A large fleet of robots need coordinated routes to execute missions efficiently. The NVIDIA cuOpt engine solves the multi-robot routing problem on the waypoint graph to produce optimized routes to multiple destinations.
Isaac Mission Dispatch, an open-sourced cloud service, is the interface between a fleet of robots and cloud services. Missions submitted to Mission Dispatch are communicated to the robot and monitored for completion.
Isaac Mission Client is an intermediate between Mission Dispatch and the robot. It publishes orders as GXF messages to the automony stack of the robot and reports the execution status to Mission Dispatch.