reward_by_max_trajectory

gistfile1.txt

prev_step_max = 0
-- reward that allows backtracking, only rewarded by getting max trajectory

function reward_by_max_trajectory()
    frame_count = frame_count + 1
    local level_done = calc_progress(data)
    local temp_progress = calc_trajectory_progress(data)

    --local reward = reward_by_ring(data) --optional, modify reward by ring behavior
    local reward = 0
    
    if (temp_progress ~= nil and temp_progress > prev_step_max) then
	    reward = (temp_progress/level_dict_len - prev_step_max/level_dict_len) * 9000
	    reward = clip(reward,-400.1,400.1)
	    prev_step_max = temp_progress
    end

    if level_done >= 1 then
	    reward = reward + (1 - clip(frame_count/frame_limit, 0, 1)) * 1000
    end

    return reward
end