Dependencies are usually a pain, especially when it involves localized hardware. Thus, instead of interacting with the robot lib api diractly, the PR enables user to run widowx as a server and client architecture. The widowx robot abstraction is kept within the docker container, and services are exposed as TCP ports. User can then run a ML/RL script in a seperate conda env (same or remote), interact with the minimal widowx client api.
The server still retain the use of previous 'widowenv' gym compliant api, which might not be the best way to do. Will require further cleanup effort in the future.
Summary:
Dependencies are usually a pain, especially when it involves localized hardware. Thus, instead of interacting with the robot lib api diractly, the PR enables user to run widowx as a server and client architecture. The widowx robot abstraction is kept within the docker container, and services are exposed as TCP ports. User can then run a ML/RL script in a seperate conda env (same or remote), interact with the minimal widowx client api.
Note: