Multi-object tracking is a crucial problem for autonomous vehicle. Most state-of-the-art approaches adopt the tracking-by-detection strategy, which is a two-step procedure consisting of the detection module and the tracking module. In this paper, we improve both steps. We improve the detection module by incorporating the temporal information, which is beneficial for detecting small objects. For the tracking module, we propose a novel compressed deep Convolutional Neural Network (CNN) feature based Correlation Filter tracker. By carefully integrating these two modules, the proposed multi-object tracking approach has the ability of re-identification (ReID) once the tracked object gets lost. Extensive experiments were performed on the KITTI and MOT2015 tracking benchmarks. Results indicate that our approach outperforms most state-of-the-art tracking approaches.

Road detection is a key task for autonomous land vehicles. Monocular vision-based road-detection algorithms are mostly based on machine learning approaches and are usually cast as classification problems. However, the pixel-wise classifiers are faced with the ambiguity caused by changes in road appearance, illumination and weather. An effective way to reduce the ambiguity is to model the contextual information with structured learning and prediction. Currently, the widely used structured prediction model in road detection is the Markov random field or conditional random field. However, the random field-based methods require additional complex optimization after pixel-wise classification, making them unsuitable for real-time applications. In this paper, we present a structured random forest-based road-detection algorithm which is capable of modelling the contextual information efficiently. By mapping the structured label space to a discrete label space, the test function of each split node can be trained in a similar way to that of the classical random forests. Structured random forests make use of the contextual information of image patches as well as the structural information of the labels to get more consistent results. Besides this benefit, by predicting a batch of pixels in a single classification, the structured random forest-based road detection can be much more efficient than the conventional pixel-wise random forest. Experimental results tested on the KITTI-ROAD dataset and data collected in typical unstructured environments show that structured random forest-based road detection outperforms the classical pixel-wise random forest both in accuracy and efficiency.

Range images are commonly used representations for 3D LiDAR point cloud in the field of autonomous driving. The approach of generating a range image is generally regarded as a standard approach. However, there do exist two different types of approaches to generating the range image: In one approach, the row of the range image is defined as the laser ID, and in the other approach, the row is defined as the elevation angle. We named the first approach Projection By Laser ID (PBID), and the second approach Projection By Elevation Angle (PBEA). Few previous works have paid attention to the difference of these two approaches. In this work, we quantitatively analyze these two different approaches. Experimental results show that the PBEA approach can obtain much smaller quantization errors than PBID, and should be the preferred choice in reconstruction-related tasks. If PBID is chosen for use in recognition-related tasks, then we have to tolerate its larger quantization error.

Pedestrian detection plays an essential role in the navigation system of autonomous vehicles. Multisensor fusion-based approaches are usually used to improve detection performance. In this study, we aimed to develop a score fusion-based pedestrian detection algorithm by integrating the data of two light detection and ranging systems (LiDARs). We first evaluated a two-stage object-detection pipeline for each LiDAR, including object proposal and fine classification. The scores from these two different classifiers were then fused to generate the result using the Bayesian rule. To improve proposal performance, we applied two features: the central points density feature, which acts as a filter to speed up the process and reduce false alarms; and the location feature, including the density distribution and height difference distribution of the point cloud, which describes an object’s profile and location in a sliding window. Extensive experiments tested in KITTI and the self-built dataset show that our method could produce highly accurate pedestrian detection results in real-time. The proposed method not only considers the accuracy and efficiency but also the flexibility for different modalities.

Trajectory prediction is an essential and challenging task for autonomous driving and mobile robots. The main difficulty is to model actor-actor interaction and actor-scene interaction. In addition, the different motion characteristics of each actor also increase the challenge of prediction. Most existing data-driven methods mainly focus on the interaction between actors but ignore the influence of their independent motion characteristics and actor-scene interaction. In this letter, we propose a multiple contextual cues integrated trajectory prediction method. Specifically, an LSTM-based encoder extracts the motion features to express the driving characteristics of each actor. Meanwhile, an attention-based graph module is applied to accurately model interaction behaviors. The scene features are extracted from high-definition vector maps by convolution neural networks. Combining these three types of attribute features, the decoder module then infers the future trajectory. We evaluate the proposed approach on two widely-used datasets, i.e. ApolloScape and Argoverse, and state-of-the-art results demonstrate the effectiveness of our approach.

Autonomous driving with artificial intelligence technology has been viewed as promising for autonomous vehicles hitting the road in the near future. In recent years, considerable progress has been made with Deep Reinforcement Learnings (DRLs) for realizing end-to-end autonomous driving. Still, driving safely and comfortably in real dynamic scenarios with DRL is nontrivial due to the reward functions being typically pre-defined with expertise. This paper proposes a human-in-the-loop DRL algorithm for learning personalized autonomous driving behavior in a progressive learning way. Specifically, a progressively optimized reward function (PORF) learning model is built and integrated into the Deep Deterministic Policy Gradient (DDPG) framework, which is called PORF-DDPG in this paper. PORF consists of two parts: the first part of the PORF is a pre-defined typical reward function on the system state, the second part is modeled as a Deep Neural Network (DNN) for representing driving adjusting intention by the human observer, which is the main contribution of this paper. The DNN-based reward model is progressively learned using the front-view images as the input and via active human supervision and intervention. The proposed approach is potentially useful for driving in dynamic constrained scenarios when dangerous collision events might occur frequently with classic DRLs. The experimental results show that the proposed autonomous driving behavior learning method exhibits online learning capability and environmental adaptability.