Adapting Autonomous Agents for Automotive Driving Games

Campodonico, Gabriele; Bellotti, Francesco; Berta, Riccardo; Capello, Alessio; Cossu, Marianna; Gloria, Alessandro De; Lazzaroni, Luca; Taccioli, Tommaso; Davio, Federico

doi:10.1007/978-3-030-92182-8_10

Cited by 6 publications

(4 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We notice that the training is quite rapid, smooth , and stable, without catastrophic forgetting [53], which frequently affects the DRL training. Low-level agent [15] Cont (20,36) Uniform(10-20) 0. 4 11 It appears that the high-level agent performs better under both the considered dimensions.…”

Section: High-level Decision Makingmentioning

confidence: 99%

“…Concerning the proposed comparison, it must be specified that the two environments are very similar to each other, but not exactly the same. Table 2 shows that an episode in Campodonico et al [15] has fewer total vehicles (EV + NPVs), but they are slightly more concentrated (density factor 0.5 vs 0.4). The model presented in [15] was specifically developed with a large number of rewards compared to the original highway-env (11 vs 3), in order to better enforce traffic law and safer behavior (e.g., penalties for right overtake, unsafe distance to the front vehicle, hazardous lane change, steering angle).…”

Section: High-level Decision Makingmentioning

confidence: 99%

“…Agents are typically trained in configurable simulation environments. Highway-env is a collection of simple 2D bird-eye view environments for training RL agents in automated driving and tactical decision-making tasks, particularly oriented to highway settings [15], [16], even if other car-related scenarios are available as well (e.g., intersection, roundabout, parking lots). Highway -env, developed by Edouard Leurent [17], is one of the environments provided within Farama Foundation Gymnasium (previously Gym [18]), an open-source Python library for developing and comparing RL algorithms by providing a standard application programming interface (API) to communicate between learning algorithms and driving environments.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Implementing Deep Reinforcement Learning (DRL)-based Driving Styles for Non-Player Vehicles

Forneris,

Pighetti,

Lazzaroni

et al. 2023

IJSG

View full text Add to dashboard Cite

We propose a new, hierarchical architecture for behavioral planning of vehicle models usable as realistic non-player vehicles in serious games related to traffic and driving. These agents, trained with deep reinforcement learning (DRL), decide their motion by taking high-level decisions, such as “keep lane”, “overtake” and “go to rightmost lane”. This is similar to a driver’s high-level reasoning and takes into account the availability of advanced driving assistance systems (ADAS) in current vehicles. Compared to a low-level decision making system, our model performs better both in terms of safety and speed. As a significant advantage, the proposed approach allows to reduce the number of training steps by more than one order of magnitude. This makes the development of new models much more efficient, which is key for implementing vehicles featuring different driving styles. We also demonstrate that, by simply tweaking the reinforcement learning (RL) reward function, it is possible to train agents characterized by different driving behaviors. We also employed the continual learning technique, starting the training procedure of a more specialized agent from a base model. This allowed significantly to reduce the number of training steps while keeping similar vehicular performance figures. However, the characteristics of the specialized agents are deeply influenced by the characteristics of the baseline agent.

show abstract

Section: High-level Decision Makingmentioning

confidence: 99%

Section: High-level Decision Makingmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Implementing Deep Reinforcement Learning (DRL)-based Driving Styles for Non-Player Vehicles

Forneris,

Pighetti,

Lazzaroni

et al. 2023

IJSG

View full text Add to dashboard Cite

show abstract

“…It must be noted that digital twins and simulation could also be used to support preparation of systems based on reinforcement learning and on other algorithms (Campodonico et al, 2021). The digital replica of our robot can be used as the training subject so save time, cost and safety concerns that would otherwise imply a training process on the real robot (Y. .…”

Section: State Of the Artmentioning

confidence: 99%

Digital twins for developing innovative industrial autonomous system

2022

Proceedings of the International Conference on Modelling and Applied Simulation MAS

View full text Add to dashboard Cite

The paper addresses the development of an innovative solution based on Autonomous Systems to operated within Industrial Plants based on the advance of technologies in this sector and investing new operational frameworks. Indeed the authors propose the adoption of digital twin approach in order to support the design of this new solution as well as on the development of new procedures and policies applied to iron & steel facilities. A real-life case study with relative system architecture and common control of the real and virtual replica is analyzed. Synthetic description of obtained results is presented.

show abstract