Discrete Policy: Learning Disentangled Action Space for Multi-Task Robotic Manipulation

Learning visuomotor policy for multi-task robotic manipulation has been a long-standing challenge for the robotics community. The difficulty lies in the diversity of action space: typically, a goal can be accomplished in multiple ways, resulting in a multimodal action distribution for a single task. The complexity of action distribution escalates as the number of tasks increases. In this work, we propose \textbf{Discrete Policy}, a robot learning method for training universal agents capable of multi-task manipulation skills. Discrete Policy employs vector quantization to map action sequences into a discrete latent space, facilitating the learning of task-specific codes. These codes are then reconstructed into the action space conditioned on observations and language instruction. We evaluate our method on both simulation and multiple real-world embodiments, including both single-arm and bimanual robot settings. We demonstrate that our proposed Discrete Policy outperforms a well-established Diffusion Policy baseline and many state-of-the-art approaches, including ACT, Octo, and OpenVLA. For example, in a real-world multi-task training setting with five tasks, Discrete Policy achieves an average success rate that is 26\% higher than Diffusion Policy and 15\% higher than OpenVLA. As the number of tasks increases to 12, the performance gap between Discrete Policy and Diffusion Policy widens to 32.5\%, further showcasing the advantages of our approach. Our work empirically demonstrates that learning multi-task policies within the latent space is a vital step toward achieving general-purpose agents.

Multi-agent Cooperative Games Using Belief Map Assisted Training

In a multi-agent system, agents share their local observations to gain global situational awareness for decision making and collaboration using a message passing system. When to send a message, how to encode a message, and how to leverage the received messages directly affect the effectiveness of the collaboration among agents. When training a multi-agent cooperative game using reinforcement learning (RL), the message passing system needs to be optimized together with the agent policies. This consequently increases the model's complexity and poses significant challenges to the convergence and performance of learning. To address this issue, we propose the Belief-map Assisted Multi-agent System (BAMS), which leverages a neuro-symbolic belief map to enhance training. The belief map decodes the agent's hidden state to provide a symbolic representation of the agent's understanding of the environment and other agent's status. The simplicity of symbolic representation allows the gathering and comparison of the ground truth information with the belief, which provides an additional channel of feedback for the learning. Compared to the sporadic and delayed feedback coming from the reward in RL, the feedback from the belief map is more consistent and reliable. Agents using BAMS can learn a more effective message passing network to better understand each other, resulting in better performance in a cooperative predator and prey game with varying levels of map complexity and compare it to previous multi-agent message passing models. The simulation results showed that BAMS reduced training epochs by 66\%, and agents who apply the BAMS model completed the game with 34.62\% fewer steps on average.

Predictive Temporal Attention on Event-based Video Stream for Energy-efficient Situation Awareness

The Dynamic Vision Sensor (DVS) is an innovative technology that efficiently captures and encodes visual information in an event-driven manner. By combining it with event-driven neuromorphic processing, the sparsity in DVS camera output can result in high energy efficiency. However, similar to many embedded systems, the off-chip communication between the camera and processor presents a bottleneck in terms of power consumption. Inspired by the predictive coding model and expectation suppression phenomenon found in human brain, we propose a temporal attention mechanism to throttle the camera output and pay attention to it only when the visual events cannot be well predicted. The predictive attention not only reduces power consumption in the sensor-processor interface but also effectively decreases the computational workload by filtering out noisy events. We demonstrate that the predictive attention can reduce 46.7% of data communication between the camera and the processor and reduce 43.8% computation activities in the processor.

SemanticSLAM: Learning based Semantic Map Construction and Robust Camera Localization

Current techniques in Visual Simultaneous Localization and Mapping (VSLAM) estimate camera displacement by comparing image features of consecutive scenes. These algorithms depend on scene continuity, hence requires frequent camera inputs. However, processing images frequently can lead to significant memory usage and computation overhead. In this study, we introduce SemanticSLAM, an end-to-end visual-inertial odometry system that utilizes semantic features extracted from an RGB-D sensor. This approach enables the creation of a semantic map of the environment and ensures reliable camera localization. SemanticSLAM is scene-agnostic, which means it doesn't require retraining for different environments. It operates effectively in indoor settings, even with infrequent camera input, without prior knowledge. The strength of SemanticSLAM lies in its ability to gradually refine the semantic map and improve pose estimation. This is achieved by a convolutional long-short-term-memory (ConvLSTM) network, trained to correct errors during map construction. Compared to existing VSLAM algorithms, SemanticSLAM improves pose estimation by 17%. The resulting semantic map provides interpretable information about the environment and can be easily applied to various downstream tasks, such as path planning, obstacle avoidance, and robot navigation. The code will be publicly available at https://github.com/Leomingyangli/SemanticSLAM

An Efficient Generalizable Framework for Visuomotor Policies via Control-aware Augmentation and Privilege-guided Distillation

Visuomotor policies, which learn control mechanisms directly from high-dimensional visual observations, confront challenges in adapting to new environments with intricate visual variations. Data augmentation emerges as a promising method for bridging these generalization gaps by enriching data variety. However, straightforwardly augmenting the entire observation shall impose excessive burdens on policy learning and may even result in performance degradation. In this paper, we propose to improve the generalization ability of visuomotor policies as well as preserve training stability from two aspects: 1) We learn a control-aware mask through a self-supervised reconstruction task with three auxiliary losses and then apply strong augmentation only to those control-irrelevant regions based on the mask to reduce the generalization gaps. 2) To address training instability issues prevalent in visual reinforcement learning (RL), we distill the knowledge from a pretrained RL expert processing low-level environment states, to the student visuomotor policy. The policy is subsequently deployed to unseen environments without any further finetuning. We conducted comparison and ablation studies across various benchmarks: the DMControl Generalization Benchmark (DMC-GB), the enhanced Robot Manipulation Distraction Benchmark (RMDB), and a specialized long-horizontal drawer-opening robotic task. The extensive experimental results well demonstrate the effectiveness of our method, e.g., showing a 17\% improvement over previous methods in the video-hard setting of DMC-GB.

SWBT: Similarity Weighted Behavior Transformer with the Imperfect Demonstration for Robotic Manipulation

Imitation learning (IL), aiming to learn optimal control policies from expert demonstrations, has been an effective method for robot manipulation tasks. However, previous IL methods either only use expensive expert demonstrations and omit imperfect demonstrations or rely on interacting with the environment and learning from online experiences. In the context of robotic manipulation, we aim to conquer the above two challenges and propose a novel framework named Similarity Weighted Behavior Transformer (SWBT). SWBT effectively learn from both expert and imperfect demonstrations without interaction with environments. We reveal that the easy-to-get imperfect demonstrations, such as forward and inverse dynamics, significantly enhance the network by learning fruitful information. To the best of our knowledge, we are the first to attempt to integrate imperfect demonstrations into the offline imitation learning setting for robot manipulation tasks. Extensive experiments on the ManiSkill2 benchmark built on the high-fidelity Sapien simulator and real-world robotic manipulation tasks demonstrated that the proposed method can extract better features and improve the success rates for all tasks. Our code will be released upon acceptance of the paper.

Cross-Modal Reasoning with Event Correlation for Video Question Answering

Video Question Answering (VideoQA) is a very attractive and challenging research direction aiming to understand complex semantics of heterogeneous data from two domains, i.e., the spatio-temporal video content and the word sequence in question. Although various attention mechanisms have been utilized to manage contextualized representations by modeling intra- and inter-modal relationships of the two modalities, one limitation of the predominant VideoQA methods is the lack of reasoning with event correlation, that is, sensing and analyzing relationships among abundant and informative events contained in the video. In this paper, we introduce the dense caption modality as a new auxiliary and distill event-correlated information from it to infer the correct answer. To this end, we propose a novel end-to-end trainable model, Event-Correlated Graph Neural Networks (EC-GNNs), to perform cross-modal reasoning over information from the three modalities (i.e., caption, video, and question). Besides the exploitation of a brand new modality, we employ cross-modal reasoning modules for explicitly modeling inter-modal relationships and aggregating relevant information across different modalities, and we propose a question-guided self-adaptive multi-modal fusion module to collect the question-oriented and event-correlated evidence through multi-step reasoning. We evaluate our model on two widely-used benchmark datasets and conduct an ablation study to justify the effectiveness of each proposed component.

Neuromorphic Online Learning for Spatiotemporal Patterns with a Forward-only Timeline

Spiking neural networks (SNNs) are bio-plausible computing models with high energy efficiency. The temporal dynamics of neurons and synapses enable them to detect temporal patterns and generate sequences. While Backpropagation Through Time (BPTT) is traditionally used to train SNNs, it is not suitable for online learning of embedded applications due to its high computation and memory cost as well as extended latency. Previous works have proposed online learning algorithms, but they often utilize highly simplified spiking neuron models without synaptic dynamics and reset feedback, resulting in subpar performance. In this work, we present Spatiotemporal Online Learning for Synaptic Adaptation (SOLSA), specifically designed for online learning of SNNs composed of Leaky Integrate and Fire (LIF) neurons with exponentially decayed synapses and soft reset. The algorithm not only learns the synaptic weight but also adapts the temporal filters associated to the synapses. Compared to the BPTT algorithm, SOLSA has much lower memory requirement and achieves a more balanced temporal workload distribution. Moreover, SOLSA incorporates enhancement techniques such as scheduled weight update, early stop training and adaptive synapse filter, which speed up the convergence and enhance the learning performance. When compared to other non-BPTT based SNN learning, SOLSA demonstrates an average learning accuracy improvement of 14.2%. Furthermore, compared to BPTT, SOLSA achieves a 5% higher average learning accuracy with a 72% reduction in memory cost.

In-Hardware Learning of Multilayer Spiking Neural Networks on a Neuromorphic Processor

Although widely used in machine learning, backpropagation cannot directly be applied to SNN training and is not feasible on a neuromorphic processor that emulates biological neuron and synapses. This work presents a spike-based backpropagation algorithm with biological plausible local update rules and adapts it to fit the constraint in a neuromorphic hardware. The algorithm is implemented on Intel Loihi chip enabling low power in-hardware supervised online learning of multilayered SNNs for mobile applications. We test this implementation on MNIST, Fashion-MNIST, CIFAR-10 and MSTAR datasets with promising performance and energy-efficiency, and demonstrate a possibility of incremental online learning with the implementation.

Neuromorphic Algorithm-hardware Codesign for Temporal Pattern Learning

Neuromorphic computing and spiking neural networks (SNN) mimic the behavior of biological systems and have drawn interest for their potential to perform cognitive tasks with high energy efficiency. However, some factors such as temporal dynamics and spike timings prove critical for information processing but are often ignored by existing works, limiting the performance and applications of neuromorphic computing. On one hand, due to the lack of effective SNN training algorithms, it is difficult to utilize the temporal neural dynamics. Many existing algorithms still treat neuron activation statistically. On the other hand, utilizing temporal neural dynamics also poses challenges to hardware design. Synapses exhibit temporal dynamics, serving as memory units that hold historical information, but are often simplified as a connection with weight. Most current models integrate synaptic activations in some storage medium to represent membrane potential and institute a hard reset of membrane potential after the neuron emits a spike. This is done for its simplicity in hardware, requiring only a "clear" signal to wipe the storage medium, but destroys temporal information stored in the neuron. In this work, we derive an efficient training algorithm for Leaky Integrate and Fire neurons, which is capable of training a SNN to learn complex spatial temporal patterns. We achieved competitive accuracy on two complex datasets. We also demonstrate the advantage of our model by a novel temporal pattern association task. Codesigned with this algorithm, we have developed a CMOS circuit implementation for a memristor-based network of neuron and synapses which retains critical neural dynamics with reduced complexity. This circuit implementation of the neuron model is simulated to demonstrate its ability to react to temporal spiking patterns with an adaptive threshold.