In many applications of social navigation, existing works have shown that predicting and reasoning about human intentions can help robotic agents make safer and more socially acceptable decisions. In this work, we study this problem for autonomous valet parking (AVP), where an autonomous vehicle ego agent must drop off its passengers, explore the parking lot, find a parking spot, negotiate for the spot with other vehicles, and park in the spot without human supervision. Specifically, we propose an AVP pipeline that selects parking spots by explicitly predicting where other agents are going to park from their motion history using learned models and probabilistic belief maps. To test this pipeline, we build a simulation environment with reactive agents and realistic modeling assumptions on the ego agent, such as occlusion-aware observations, and imperfect trajectory prediction. Simulation experiments show that our proposed method outperforms existing works that infer intentions from future predicted motion or embed them implicitly in end-to-end models, yielding better results in prediction accuracy, social acceptance, and task completion. Our key insight is that, in parking, where driving regulations are more lax, explicit intention prediction is crucial for reasoning about diverse and ambiguous long-term goals, which cannot be reliably inferred from short-term motion prediction alone, but can be effectively learned from motion history.
Figures & tables
Fig. 1: A flow chart of our method, with each component marked by its corresponding section number. At each timestep, we combine information from the current observation and a probabilistic belief map of where cars are parked to reconstruct the bird’s-eye view (BEV) map of other agents. This enables the use of a learned intention model, which we use to make safe and socially acceptable parking decisions.
Fig. 2: An overview of the methods under comparison, with legend shown on top right. In this scenario, the dynamic vehicle (other agent) is backing into a spot after moving forward to adjust its angle (a “right-south-up” maneuver [ 19 ] ). Our proposed method is the only one to explicitly and correctly predicted this behavior.
Fig. 3: Setup of our experiment at t=0 . At each iteration, some of the bottom 10 spots are randomly chosen to be vacant. Then, 1 or 2 dynamic vehicles would spawn, with pots being their goal. There are always 2 vacant spots on the sides in case the ego failed to compete.
Fig. 4: Example trajectory predictions (yellow outline) conditioned on the predicted intention (green spot) which we compared in our experiment.
Success
Social Acceptance
Computation Time (s) ↓
Parking
Accuracy
Method
Rate
Spots
Interrupted
Spot
Path
Time
minADE
minFDE
(%) ↑
Stolen (%) ↓
Timesteps ↓
Selection
Planning
(s) ↓
(m) ↓
(m) ↓
Non-Reactive Agents
Cubic
99.8
0.0
0.000
0.0052
1.2812
20.1
2.02
2.40
Bézier (Ours)
± 0.000
± 0.0026
± 0.2516
± 9.6
± 1.22
± 1.89
Constant
92.0
0.2
0.000
0.0058
1.2848
19.3
3.09
5.59
TABLE I: Results of our proposed method with different trajectory prediction methods, a method that predicts intentions from trajectory prediction [ 20 ] , and a method that reasons with intentions implicitly [ 23 ] for 500 parking simulations with non-interactive agents and 500 simulations with interactive agents. The best results in each evaluation metric ( ↑ : higher is better, ↓ : lower is better) are highlighted.
Automated parking is a challenging operational domain for advanced driver assistance systems, requiring robust scene understanding and interaction reasoning. The key challenge is twofold: (i) predict multiple plausible ego intentions according to context and (ii) for each intention, predict the joint responses of surrounding agents, enabling effective what-if decision-making. However, existing methods often fall short, typically treating these interdependent problems in isolation. We propose ParkDiffusion++, which jointly learns a multi-modal ego intention predictor and an ego-conditioned multi-agent joint trajectory predictor for automated parking. Our approach makes several key contributions. First, we introduce an ego intention tokenizer that predicts a small set of discrete endpoint intentions from agent histories and vectorized map polylines. Second, we perform ego-intention-conditioned joint prediction, yielding socially consistent predictions of the surrounding agents for each possible ego intention. Third, we employ a lightweight safety-guided denoiser with different constraints to refine joint scenes during training, thus improving accuracy and safety. Fourth, we propose counterfactual knowledge distillation, where an EMA teacher refined by a frozen safety-guided denoiser provides pseudo-targets that capture how agents react to alternative ego intentions. Extensive evaluations demonstrate that ParkDiffusion++ achieves state-of-the-art performance on the Dragon Lake Parking (DLP) dataset and the Intersections Drone (inD) dataset. Importantly, qualitative what-if visualizations show that other agents react appropriately to different ego intentions.
Jiarong Wei, Anna Rehr, Christian Feist +1
Department of Computer Science, University of Freiburg, Germany. · CARIAD SE, Germany.
Existing methods in Autonomous Valet Parking (AVP) typically rely on pre-built maps, which severely restricts their scalability to unseen environments and open-vocabulary targets. Inspired by the application of Vision-Language Models (VLMs) in Vision-Language Navigation (VLN) tasks, we propose VLN-AVP, a zero-shot navigation framework for AVP tasks. By combining the precise spatial perception of a Bird's-Eye-View (BEV) model with the general intelligence of VLMs, our framework 1) eliminates the dependency on pre-built maps, 2) interprets semantic environmental contexts in parking scenarios, and 3) enables intuitive navigation following natural language instructions. Specifically, we introduce a hybrid memory system: a short-term perception memory tracks semantic visual cues to address the limitations of VLM's single-frame reasoning in existing methods, while a long-term topological memory facilitates stable policy learning from past experiences. To bridge the gap in existing benchmarks, we also present the VLN-AVP dataset and benchmark. Featuring 10 high-fidelity parking scenes and over 1,000 navigation episodes, it has the largest number of garage scenes to date and is the first VLN benchmark for underground parking. Extensive experiments demonstrate that in simulation, our method achieves an over 25% improvement in success rate compared to VLN methods and an over 15% improvement compared to other autonomous driving methods. Furthermore, it attains a leading success rate in real-world vehicle experiments, proving its practical feasibility.
Autonomous parking fundamentally differs from on-road driving due to its frequent direction changes and complex maneuvering requirements. However, existing End-to-End (E2E) planning methods often simplify the parking task into a geometric path regression problem, neglecting explicit modeling of the vehicle's kinematic state. This "dimensionality deficiency" easily leads to physically infeasible trajectories and deviates from real human driving behavior, particularly at critical gear-shift points in multi-shot parking scenarios. In this paper, we propose SunnyParking, a novel dual-branch E2E architecture that achieves motion state awareness by jointly predicting spatial trajectories and discrete motion state sequences (e.g., forward/reverse). Additionally, we introduce a Fourier feature-based representation of target parking slots to overcome the resolution limitations of traditional bird's-eye view (BEV) approaches, enabling high-precision target interactions. Experimental results demonstrate that our framework generates more robust and human-like trajectories in complex multi-shot parking scenarios, while significantly improving gear-shift point localization accuracy compared to state-of-the-art methods. We open-source a new parking dataset of the CARLA simulator, specifically designed to evaluate full prediction capabilities under complex maneuvers.
Jishu Miao, Han Chen, Jiankun Zhai +4
Jishu Miao, Tsubasa Hirakawa, Takayoshi Yamashita, Hironobu Fujiyoshi are with Chubu University, Kasugai, Japan. · Han Chen is with Tongji University, Shanghai, China. · Jiankun Zhai, Qi Liu are independent researchers.