Agents must be able to adapt quickly as an environment changes. We find that existing model-based reinforcement learning agents are unable to do this well, in part because of how they use past experiences to train their world model. Here, we present Curious Replay -- a form of prioritized experience replay tailored to model-based agents through use of a curiosity-based priority signal. Agents using Curious Replay exhibit improved performance in an exploration paradigm inspired by animal behavior and on the Crafter benchmark. DreamerV3 with Curious Replay surpasses state-of-the-art performance on Crafter, achieving a mean score of 19.4 that substantially improves on the previous high score of 14.5 by DreamerV3 with uniform replay, while also maintaining similar performance on the Deepmind Control Suite. Code for Curious Replay is available at https://github.com/AutonomousAgentsLab/curiousreplay

Motion forecasting is a key module in an autonomous driving system. Due to the heterogeneous nature of multi-sourced input, multimodality in agent behavior, and low latency required by onboard deployment, this task is notoriously challenging. To cope with these difficulties, this paper proposes a novel agent-centric model with anchor-informed proposals for efficient multimodal motion prediction. We design a modality-agnostic strategy to concisely encode the complex input in a unified manner. We generate diverse proposals, fused with anchors bearing goal-oriented scene context, to induce multimodal prediction that covers a wide range of future trajectories. Our network architecture is highly uniform and succinct, leading to an efficient model amenable for real-world driving deployment. Experiments reveal that our agent-centric network compares favorably with the state-of-the-art methods in prediction accuracy, while achieving scene-centric level inference latency.

Various methods for Multi-Agent Reinforcement Learning (MARL) have been developed with the assumption that agents' policies are based on accurate state information. However, policies learned through Deep Reinforcement Learning (DRL) are susceptible to adversarial state perturbation attacks. In this work, we propose a State-Adversarial Markov Game (SAMG) and make the first attempt to investigate the fundamental properties of MARL under state uncertainties. Our analysis shows that the commonly used solution concepts of optimal agent policy and robust Nash equilibrium do not always exist in SAMGs. To circumvent this difficulty, we consider a new solution concept called robust agent policy, where agents aim to maximize the worst-case expected state value. We prove the existence of robust agent policy for finite state and finite action SAMGs. Additionally, we propose a Robust Multi-Agent Adversarial Actor-Critic (RMA3C) algorithm to learn robust policies for MARL agents under state uncertainties. Our experiments demonstrate that our algorithm outperforms existing methods when faced with state perturbations and greatly improves the robustness of MARL policies. Our code is public on https://songyanghan.github.io/what_is_solution/.

Probabilistic user modeling is essential for building machine learning systems in the ubiquitous cases with humans in the loop. However, modern advanced user models, often designed as cognitive behavior simulators, are incompatible with modern machine learning pipelines and computationally prohibitive for most practical applications. We address this problem by introducing widely-applicable differentiable surrogates for bypassing this computational bottleneck; the surrogates enable computationally efficient inference with modern cognitive models. We show experimentally that modeling capabilities comparable to the only available solution, existing likelihood-free inference methods, are achievable with a computational cost suitable for online applications. Finally, we demonstrate how AI-assistants can now use cognitive models for online interaction in a menu-search task, which has so far required hours of computation during interaction.

Chinese Premier Li Qiang has called for more "communication and exchange" to avoid misunderstanding in his remarks at the opening of this year's'Summer Davos' in Tianjin, the first in-person event in three years following the COVID-19 pandemic. The three-day summit, which got under way on Tuesday, is hosted by the World Economic Forum but will focus heavily on China's place in the world and concerns about how the global economy can move forward in an increasingly fractious world, according to the agenda. Li told delegates it was time to support globalisation and deeper economic cooperation. "In the West, some people are hyping up what is called'cutting reliance and de-risking'," Li said. "These two concepts… are a false proposition, because the development of economic globalisation is such that the world economy has become a common entity in which you and I are both intermingled. The economies of many countries are blended with each other, rely on each other, make accomplishments because of one another and develop together. This is actually a good thing, not a bad thing."

Training AI with strong and rich strategies in multi-agent environments remains an important research topic in Deep Reinforcement Learning (DRL). The AI's strength is closely related to its diversity of strategies, and this relationship can guide us to train AI with both strong and rich strategies. To prove this point, we propose Diversity is Strength (DIS), a novel DRL training framework that can simultaneously train multiple kinds of AIs. These AIs are linked through an interconnected history model pool structure, which enhances their capabilities and strategy diversities. We also design a model evaluation and screening scheme to select the best models to enrich the model pool and obtain the final AI. The proposed training method provides diverse, generalizable, and strong AI strategies without using human data. We tested our method in an AI competition based on Google Research Football (GRF) and won the 5v5 and 11v11 tracks. The method enables a GRF AI to have a high level on both 5v5 and 11v11 tracks for the first time, which are under complex multi-agent environments. The behavior analysis shows that the trained AI has rich strategies, and the ablation experiments proved that the designed modules benefit the training process.

We present a computational model for a symbol emergence system that enables the emergence of lexical knowledge with combinatoriality among agents through a Metropolis-Hastings naming game and cross-situational learning. Many computational models have been proposed to investigate combinatoriality in emergent communication and symbol emergence in cognitive and developmental robotics. However, existing models do not sufficiently address category formation based on sensory-motor information and semiotic communication through the exchange of word sequences within a single integrated model. Our proposed model facilitates the emergence of lexical knowledge with combinatoriality by performing category formation using multimodal sensory-motor information and enabling semiotic communication through the exchange of word sequences among agents in a unified model. Furthermore, the model enables an agent to predict sensory-motor information for unobserved situations by combining words associated with categories in each modality. We conducted two experiments with two humanoid robots in a simulated environment to evaluate our proposed model. The results demonstrated that the agents can acquire lexical knowledge with combinatoriality through interpersonal cross-situational learning based on the Metropolis-Hastings naming game and cross-situational learning. Furthermore, our results indicate that the lexical knowledge developed using our proposed model exhibits generalization performance for novel situations through interpersonal cross-modal inference.

In the field of Geriatronics, enabling effective and transparent communication between humans and robots is crucial for enhancing the acceptance and performance of assistive robots. Our early-stage research project investigates the potential of language-based modulation as a means to improve human-robot interaction. We propose to explore real-time modulation during task execution, leveraging language cues, visual references, and multimodal inputs. By developing transparent and interpretable methods, we aim to enable robots to adapt and respond to language commands, enhancing their usability and flexibility. Through the exchange of insights and knowledge at the workshop, we seek to gather valuable feedback to advance our research and contribute to the development of interactive robotic systems for Geriatronics and beyond.

Text-based games are a popular testbed for language-based reinforcement learning (RL). In previous work, deep Q-learning is commonly used as the learning agent. Q-learning algorithms are challenging to apply to complex real-world domains due to, for example, their instability in training. Therefore, in this paper, we adapt the soft-actor-critic (SAC) algorithm to the text-based environment. To deal with sparse extrinsic rewards from the environment, we combine it with a potential-based reward shaping technique to provide more informative (dense) reward signals to the RL agent. We apply our method to play difficult text-based games. The SAC method achieves higher scores than the Q-learning methods on many games with only half the number of training steps. This shows that it is well-suited for text-based games. Moreover, we show that the reward shaping technique helps the agent to learn the policy faster and achieve higher scores. In particular, we consider a dynamically learned value function as a potential function for shaping the learner's original sparse reward signals.

Unmanned aerial vehicles (UAVs) are increasingly deployed to provide wireless connectivity to static and mobile ground users in situations of increased network demand or points of failure in existing terrestrial cellular infrastructure. However, UAVs are energy-constrained and experience the challenge of interference from nearby UAV cells sharing the same frequency spectrum, thereby impacting the system's energy efficiency (EE). Recent approaches focus on optimising the system's EE by optimising the trajectory of UAVs serving only static ground users and neglecting mobile users. Several others neglect the impact of interference from nearby UAV cells, assuming an interference-free network environment. Despite growing research interest in decentralised control over centralised UAVs' control, direct collaboration among UAVs to improve coordination while optimising the systems' EE has not been adequately explored. To address this, we propose a direct collaborative communication-enabled multi-agent decentralised double deep Q-network (CMAD-DDQN) approach. The CMAD-DDQN is a collaborative algorithm that allows UAVs to explicitly share their telemetry via existing 3GPP guidelines by communicating with their nearest neighbours. This allows the agent-controlled UAVs to optimise their 3D flight trajectories by filling up knowledge gaps and converging to optimal policies. Simulation results show that the proposed approach outperforms existing baselines in terms of maximising the systems' EE without degrading coverage performance in the network. The CMAD-DDQN approach outperforms the MAD-DDQN that neglects direct collaboration among UAVs, the multi-agent deep deterministic policy gradient (MADDPG) and random policy approaches that consider a 2D UAV deployment design while neglecting interference from nearby UAV cells by about 15%, 65% and 85%, respectively.