Abstract:Predicting ego vehicle trajectories remains a critical challenge, especially in urban and dense areas due to the unpredictable behaviours of other vehicles and pedestrians. Multimodal trajectory prediction enhances decision-making by considering multiple possible future trajectories based on diverse sources of environmental data. In this approach, we leverage ResNet-50 to extract image features from high-definition map data and use IMU sensor data to calculate speed, acceleration, and yaw rate. A temporal probabilistic network is employed to compute potential trajectories, selecting the most accurate and highly probable trajectory paths. This method integrates HD map data to improve the robustness and reliability of trajectory predictions for autonomous vehicles.

What problem does this paper attempt to address?

This paper attempts to address the challenge of predicting the trajectory of an autonomous vehicle (i.e., the ego - vehicle) in complex urban environments. Specifically, the paper focuses on how to utilize high - definition maps (HD Map) and sensor data to improve the accuracy and reliability of trajectory prediction, especially in cases where the behavior of pedestrians and other vehicles is unpredictable. ### Main Problems 1. **Trajectory Prediction in Complex Environments**: - In urban and dense areas, due to the unpredictable behavior of other vehicles and pedestrians, the trajectory prediction of the ego - vehicle becomes particularly difficult. - Traditional physics - based methods perform well in short - term prediction but not in long - term prediction because they cannot handle the influence of external factors (such as weather, traffic conditions, and driver intentions) well. 2. **Multi - modal Trajectory Prediction**: - A single trajectory prediction method may not be sufficient to deal with complex traffic scenarios, so multiple possible future trajectories need to be considered. - Multi - modal prediction can consider different behavior patterns (such as changes in speed and direction), thereby improving the robustness of prediction. 3. **Fusion of Multiple Data Sources**: - High - definition maps provide detailed environmental information, such as lane markings, traffic signs, and road shapes. - Sensor data (such as the speed, acceleration, and yaw rate provided by IMU sensors) provide real - time state information of the vehicle. - How to effectively fuse these data from different sources is a key issue. ### Solutions The paper proposes a model named MapsTP, which solves the above problems in the following ways: - **Feature Extraction**: Use ResNet - 50 to extract features from high - definition map images and combine IMU sensor data to calculate the vehicle's speed, acceleration, and yaw rate. - **Spatio - Temporal Probability Network (STPN)**: Calculate multiple potential trajectories through a spatio - temporal probability network, assign probabilities to each trajectory, and select the most likely path. - **Multi - modal Prediction**: Consider multiple possible future trajectories to deal with uncertainty and complex behavior patterns. ### Summary The main objective of the paper is to develop a model that can accurately predict the trajectory of the ego - vehicle by fusing high - definition map and sensor data, thereby improving the safety and efficiency of autonomous vehicles in complex urban environments.

MapsTP: HD Map Images Based Multimodal Trajectory Prediction for Automated Vehicles

Enhanced Multimodal Trajectory Prediction for Autonomous Vehicles Using Advanced Diffusion Model Techniques

Multimodal Trajectory Predictions for Autonomous Driving using Deep Convolutional Networks

BEVSeg2TP: Surround View Camera Bird's-Eye-View Based Joint Vehicle Segmentation and Ego Vehicle Trajectory Prediction

Video Killed the HD-Map: Predicting Multi-Agent Behavior Directly From Aerial Images

Trajectory Prediction-Based Local Spatio-Temporal Navigation Map for Autonomous Driving in Dynamic Highway Environments

Map-Adaptive Multimodal Trajectory Prediction via Intention-Aware Unimodal Trajectory Predictors

Multimodal Trajectory Prediction for Diverse Vehicle Types in Autonomous Driving with Heterogeneous Data and Physical Constraints

Trajectory Prediction for Autonomous Driving with Topometric Map

Trajectory Prediction for Human-driven Vehicles Based on a Combined Method of Deep Neural Network and Driving Risk Map

Multiagent Multimodal Trajectory Prediction in Urban Traffic Scenarios Using a Neural Network-Based Solution

MFTraj: Map-Free, Behavior-Driven Trajectory Prediction for Autonomous Driving

VisionNet: A Drivable-space-based Interactive Motion Prediction Network for Autonomous Driving

A CNN-LSTM Based Model to Predict Trajectory of Human-Driven Vehicle

A Cognitive-Based Trajectory Prediction Approach for Autonomous Driving

Physically constrained short-term vehicle trajectory forecasting with naive semantic maps

Multi-modal Trajectory Prediction for Autonomous Driving with Semantic Map and Dynamic Graph Attention Network

Predicting Maps Using In-Vehicle Cameras for Data-Driven Intelligent Transport

Shared Cross-Modal Trajectory Prediction for Autonomous Driving

A Probabilistic Architecture of Long-Term Vehicle Trajectory Prediction for Autonomous Driving

Graph-Based Interaction-Aware Multimodal 2D Vehicle Trajectory Prediction using Diffusion Graph Convolutional Networks