TSP = Travelling Salesman Problem (https://en.wikipedia.org/wiki/Travelling_salesman_problem)

PPO = Proximal Policy Optimisation, a reinforcement learning algorithm (https://en.wikipedia.org/wiki/Proximal_Policy_Optimization)

Also compare with LKH3 which seems much faster and closer to optimal.

Sorry if I am harsh, but a 1200 node tsp problem is a toy problem. We can find proven optimal solutions to these in a fraction of the time you spent.

RL is probably best suited for uncertainty infected instances.