Conditional Imitation learning is a common and effective approach to train autonomous driving agents. However, two issues limit the full potential of this approach: (i) the inertia problem, a special case of causal confusion where the agent mistakenly correlates low speed with no acceleration, and (ii) low correlation between offline and online performance due to the accumulation of small errors that brings the agent in a previously unseen state. Both issues are critical for state-aware models, yet informing the driving agent of its internal state as well as the state of the environment is of crucial importance. In this paper we propose a multi-task learning agent based on a multi-stage vision transformer with state token propagation. We feed the state of the vehicle along with the representation of the environment as a special token of the transformer and propagate it throughout the network. This allows us to tackle the aforementioned issues from different angles: guiding the driving policy with learned stop/go information, performing data augmentation directly on the state of the vehicle and visually explaining the model's decisions. We report a drastic decrease in inertia and a high correlation between offline and online metrics.

Cultrera, L., Becattini, F., Seidenari, L., Pala, P., Bimbo, A.D. (2024). Addressing Limitations of State-Aware Imitation Learning for Autonomous Driving. IEEE TRANSACTIONS ON INTELLIGENT VEHICLES, 10 [10.1109/TIV.2023.3336063].

Addressing Limitations of State-Aware Imitation Learning for Autonomous Driving

Becattini, Federico
;
2024-01-01

Abstract

Conditional Imitation learning is a common and effective approach to train autonomous driving agents. However, two issues limit the full potential of this approach: (i) the inertia problem, a special case of causal confusion where the agent mistakenly correlates low speed with no acceleration, and (ii) low correlation between offline and online performance due to the accumulation of small errors that brings the agent in a previously unseen state. Both issues are critical for state-aware models, yet informing the driving agent of its internal state as well as the state of the environment is of crucial importance. In this paper we propose a multi-task learning agent based on a multi-stage vision transformer with state token propagation. We feed the state of the vehicle along with the representation of the environment as a special token of the transformer and propagate it throughout the network. This allows us to tackle the aforementioned issues from different angles: guiding the driving policy with learned stop/go information, performing data augmentation directly on the state of the vehicle and visually explaining the model's decisions. We report a drastic decrease in inertia and a high correlation between offline and online metrics.
2024
Cultrera, L., Becattini, F., Seidenari, L., Pala, P., Bimbo, A.D. (2024). Addressing Limitations of State-Aware Imitation Learning for Autonomous Driving. IEEE TRANSACTIONS ON INTELLIGENT VEHICLES, 10 [10.1109/TIV.2023.3336063].
File in questo prodotto:
File Dimensione Formato  
Addressing_Limitations_of_State-Aware_Imitation_Learning_for_Autonomous_Driving.pdf

accesso aperto

Tipologia: Pre-print
Licenza: Creative commons
Dimensione 4.41 MB
Formato Adobe PDF
4.41 MB Adobe PDF Visualizza/Apri

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11365/1255275