Model analysis report for

• ▶
  root: {} 10 keys
  • report_name: 'rldriving_driving.analyse_lat_cheat_rl'
  • route_list: '/home/batman/xx/routes/default_test_2k'
  • output_dir: null
  • save_tmp: false
  • read_only: false
  • limit: 10000
  • use_virtual_webgl: false
  • format: 'HTML'
  • ▶
    miniray: {} 8 keys
    • priority: 10
    • job_name: 'rldriving_driving.analyse_lat_cheat_rl'
    • queue_name: 'remote2'
    • submit_host: 'redis.comma.internal'
    • result_host: 'data-ram.comma.internal'
    • no_unique_id: false
    • codedir: 'caches/workstation-yassine_2025-06-17_09-59-38'
    • force_local: false
  • ▶
    rollout: {} 10 keys
    • type: 'TinySimRL'
    • nowindow: false
    • debug: false
    • debug_render: false
    • save_observation_keys: []
    • output_dir: null
    • rerun: false
    • ▶
      env: {} 30 keys
      • debug_window_size: [512, 512]
      • ▶
        ep: {} 7 keys
        • segment: ''
        • augment_desire: false
        • depth_type: 'EASYDEPTH'
        • calibration_source: 'LOG'
        • pipeline_dir: 'http://data-ssd.comma.life/runner/training_2024_06'
        • feature_dir: null
        • ▶
          input_paths: {} 0 keys
      • start: null
      • end: null
      • env: 'xx.arena.envs.tinysim_env.TinySimEnv'
      • get_imgs: true
      • allow_resets: true
      • zero_desire: true
      • fps: 20
      • allow_lat: true
      • allow_long: false
      • min_hist: 40
      • compressor_model: null
      • compressor_mode: 'NONE'
      • quantized_compressor_latents: false
      • yolo_model: false
      • lpips_model: false
      • sharpness_metric: true
      • future_observation_range_seconds: null
      • max_x_error: 4.0
      • max_y_error: 3.0
      • max_yaw_error: 0.2
      • reset_every: null
      • noise_model: 'REALISTIC'
      • photo_noise_model: 'NONE'
      • augment_mode: 'DEPTH'
      • inpaint_mode: 'NONE'
      • avoid_hood: false
      • cache_granularity: true
      • worldmodel: null
    • ▶
      agent: {} 12 keys
      • supercombo: '967279c1-7d3c-4463-9d35-58e0311a5f57/700'
      • restrict_outputs: null
      • exclude_outputs: ['imgs']
      • calibrate_speed: false
      • use_allow_throttle: false
      • use_prev_desired_curv: true
      • model_trained_fps: 5
      • use_worldmodel_as_agent: false
      • use_raw_plan_lat: false
      • use_raw_plan_long: false
      • desired_curv_smooth_seconds: false
      • desired_accel_smooth_seconds: false
    • ▶
      shadow_agent: {} 0 keys

Plan analysis

All segments

Metrics

passed :1274 / 1658 aka 76.840%

better :939 / 1658 aka 56.634%

best (0.5m):436 / 1658 aka 26.297%

global mae x: 0.168

global mae y: 0.264

global mae y of lanecenter (when detected) in GT: 0.209

global mae y of lanecenter (when detected) in Sim: 0.254

avg resets: 0.361

Mean abs curvature: 0.215 deg/m

Mean abs curvature rate: 0.306 deg/ms

Mean abs lateral accel: 0.369 m/(ss)

Mean abs lateral accel GT: 0.336 m/(ss)

Mean abs lateral jerk: 0.497 m/(sss)

Mean abs lateral jerk GT: 0.272 m/(sss)

Easy straight

Metrics

passed :508 / 510 aka 99.608%

better :493 / 510 aka 96.667%

best (0.5m):294 / 510 aka 57.647%

global mae x: 0.106

global mae y: 0.175

global mae y of lanecenter (when detected) in GT: 0.161

global mae y of lanecenter (when detected) in Sim: 0.182

avg resets: 0.016

Mean abs curvature: 0.048 deg/m

Mean abs curvature rate: 0.087 deg/ms

Mean abs lateral accel: 0.174 m/(ss)

Mean abs lateral accel GT: 0.130 m/(ss)

Mean abs lateral jerk: 0.305 m/(sss)

Mean abs lateral jerk GT: 0.171 m/(sss)

Low speed

Metrics

passed :295 / 347 aka 85.014%

better :193 / 347 aka 55.620%

best (0.5m):75 / 347 aka 21.614%

global mae x: 0.228

global mae y: 0.278

global mae y of lanecenter (when detected) in GT: 0.266

global mae y of lanecenter (when detected) in Sim: 0.271

avg resets: 0.473

Mean abs curvature: 0.418 deg/m

Mean abs curvature rate: 0.609 deg/ms

Mean abs lateral accel: 0.312 m/(ss)

Mean abs lateral accel GT: 0.305 m/(ss)

Mean abs lateral jerk: 0.439 m/(sss)

Mean abs lateral jerk GT: 0.228 m/(sss)

No desire segments

Metrics

passed :921 / 938 aka 98.188%

better :824 / 938 aka 87.846%

best (0.5m):423 / 938 aka 45.096%

global mae x: 0.125

global mae y: 0.196

global mae y of lanecenter (when detected) in GT: 0.180

global mae y of lanecenter (when detected) in Sim: 0.192

avg resets: 0.065

Mean abs curvature: 0.119 deg/m

Mean abs curvature rate: 0.160 deg/ms

Mean abs lateral accel: 0.318 m/(ss)

Mean abs lateral accel GT: 0.284 m/(ss)

Mean abs lateral jerk: 0.387 m/(sss)

Mean abs lateral jerk GT: 0.236 m/(sss)

No desire turns (max lat accel > 2.5 m/ss)

Metrics

passed :124 / 132 aka 93.939%

better :74 / 132 aka 56.061%

best (0.5m):17 / 132 aka 12.879%

global mae x: 0.218

global mae y: 0.258

global mae y of lanecenter (when detected) in GT: 0.260

global mae y of lanecenter (when detected) in Sim: 0.233

avg resets: 0.189

Mean abs curvature: 0.286 deg/m

Mean abs curvature rate: 0.325 deg/ms

Mean abs lateral accel: 0.665 m/(ss)

Mean abs lateral accel GT: 0.649 m/(ss)

Mean abs lateral jerk: 0.619 m/(sss)

Mean abs lateral jerk GT: 0.398 m/(sss)

Left lane change segments

36/120 aka 30.00%

better: 13/120 aka 10.83%

best (0.5m): 3/120 aka 2.50%

mean max error C3: 2.150037

mean max error C2: 2.361451

global mae: 0.345811

avg resets: 0.366667

Right lane change segments

42/150 aka 28.00%

better: 11/150 aka 7.33%

best (0.5m): 4/150 aka 2.67%

mean max error C3: 2.284184

mean max error C2: 2.419934

global mae: 0.345036

avg resets: 0.613333

Left turn segments

68/79 aka 86.08%

better: 23/79 aka 29.11%

best (0.5m): 0/79 aka 0.00%

mean max error C3: 1.289036

mean max error C2: 1.489890

global mae: 0.291154

avg resets: 0.822785

Right turn segments

72/91 aka 79.12%

better: 35/91 aka 38.46%

best (0.5m): 2/91 aka 2.20%

mean max error C3: 1.324712

mean max error C2: 1.380363

global mae: 0.298632

avg resets: 0.714286

Left keep segments

20/33 aka 60.61%

better: 9/33 aka 27.27%

best (0.5m): 1/33 aka 3.03%

mean max error C3: 1.573230

mean max error C2: 1.833280

global mae: 0.290920

avg resets: 0.333333

Right keep segments

20/26 aka 76.92%

better: 9/26 aka 34.62%

best (0.5m): 3/26 aka 11.54%

mean max error C3: 1.956904

mean max error C2: 1.248492

global mae: 0.301362

avg resets: 0.192308

Plan y error in turns (> 6°/s)

Mean path error : -0.006 meters

Mean absolute path error : 0.448 meters

Turn bias test (y_pos at 50m > 3m)

got 29069 left samples

got 30815 right samples

mean left gt: -0.020020194

mean right gt: 0.012990205

mean turn lanecenter error (GT pos, GT ll): left = 0.21726641058921814, right = -0.24762769043445587

mean turn lanecenter error (sim pos, GT ll): left = 0.2224971204996109, right = -0.2433958202600479

mean turn lanecenter error (GT pos, model ll): left = 0.19507455825805664, right = -0.2593687176704407

mean turn lanecenter error (sim pos, model ll): left = 0.17884349822998047, right = -0.23043349385261536

Mean abs curvature: 0.213776 deg/m

Mean abs curvature rate: 0.205523 deg/ms

Mean gt pos in left and right turns

Lane position frequency-domain analysis

Mean frequency: 0.174532 Hz, period 5.729611 s

Mean amplitude: 0.131274 m

Mean ping-pong significance (frequency x amplitude): 2.242652 %