MuJoCo
PulseAugur coverage of MuJoCo — every cluster mentioning MuJoCo across labs, papers, and developer communities, ranked by signal.
9 day(s) with sentiment data
-
FEAGINE unveils flexible robot bodies and cross-embodiment AI model
FEAGINE, a robotics company founded by former DJI engineer Peng Rui, is developing a new approach to embodied AI that moves beyond the humanoid form. Instead of replicating human bodies, FEAGINE is designing flexible, b…
-
New DClamp-PPO algorithm enhances reinforcement learning by penalizing 'wrong' direction updates
Researchers have introduced Directional-Clamp PPO (DClamp-PPO), a novel algorithm designed to enhance the performance of Proximal Policy Optimization (PPO) in deep reinforcement learning. DClamp-PPO addresses a key limi…
-
New MINT method optimizes multi-timescale interventions under constraints
Researchers have developed a new method called MINT (Multi-timescale Intervention Network Training) to address sequential decision problems where interventions can have immediate or persistent effects. MINT uses an augm…
-
New method distills nonlinear dynamics into linear state-space models
Researchers have developed a novel pipeline for learning linear state-space models from nonlinear dynamical systems. This method, termed Spectral Distillation, uses Observation Spectral Filtering (OSF) to first learn an…
-
New framework and simulation tool tackle inevitable robot failures
Researchers have developed a new safety framework to address inevitable failures in service robots operating in human-shared environments. This framework quanties the probability and severity of negative interactions du…
-
New SO-OPF method precisely analyzes vision encoder changes
Researchers have developed a new method called Support Operation Factorization (SO-OPF) to analyze frozen vision encoders, aiming to precisely identify what changes and where within the encoder's operations. This techni…
-
Ambiguous API requests like "MiMo" require structured identification to prevent integration errors
The article discusses the ambiguity of API requests when a single name, like "MiMo," can refer to multiple unrelated products. It highlights a common intake process error where a vague request for "MiMo AI API" is mista…
-
New Catheter Control Method Prioritizes Safety and Tracking
Researchers have developed a new method for controlling steerable catheters, focusing on the dynamics of interaction between the catheter and tissue. The approach uses a partial-physics feedforward to cancel reliable be…
-
New RL method PGTT enhances legged robot terrain traversal
Researchers have developed a new reinforcement learning approach called Phase-Guided Terrain Traversal (PGTT) for legged robots. This method uses reward shaping to enforce gait structure, allowing policies to operate di…
-
New PAMD method enhances visual reinforcement learning algorithms
Researchers have introduced PAMD, a novel Pairwise Adaptive Mahalanobis Distance method designed to improve visual reinforcement learning algorithms. This new approach parameterizes a positive-definite, pair-conditioned…
-
New benchmark for robotic arm reach-avoid task using DRL
Researchers have developed a new benchmark for the reach-avoid task in robotics, utilizing the MuJoCo MJX physics engine and the Brax library for parallelized simulation and reinforcement learning. This benchmark aims t…
-
AI tactile tech firm Tashan Technology secures hundreds of millions in new funding · 1 source tracked
Tashan Technology, a Beijing-based company specializing in AI tactile perception, has secured several hundred million yuan in Series B funding. The investment, led by Junshi Electronics and including other industry play…
-
New MAGIK framework enables zero-shot knowledge transfer in RL agents
Researchers have developed MAGIK, a novel framework designed to enhance knowledge transfer in reinforcement learning (RL) agents. This system enables RL agents to apply knowledge from previously learned tasks to new, an…
-
Fly brain topology inspires robust neural network for robot navigation
Researchers have developed a new recurrent neural network called FLYNN, which is directly modeled after the neural architecture of a fruit fly's brain. This network demonstrates robust navigation capabilities in simulat…
-
New research tackles LLM alignment, safety, and optimization challenges
Researchers are exploring new methods to improve the alignment and reliability of large language models (LLMs). One study identifies a vulnerability in byte-pair encoding (BPE) tokenization that can be exploited to bypa…
-
New AIDA framework improves visual reinforcement learning with limited data
Researchers have developed AIDA (Adaptive Imagination for Domain Adaptation), a novel framework designed to improve visual reinforcement learning in scenarios with limited target data. This approach addresses the sim-to…
-
New research revisits action factorization for complex RL spaces · 2 sources tracked
A new research paper explores methods for handling complex action spaces in reinforcement learning, particularly those that combine discrete and continuous actions. The study analyzes various factorization techniques ac…
-
New open-source simulator MuJoFil targets high-fidelity vision RL training
A new open-source simulator called MuJoFil has been developed, aiming to address limitations in existing tools like MuJoCo for high-fidelity vision reinforcement learning (RL) training. MuJoFil combines Nvidia's GPU-nat…
-
New research unifies PPO-Clip and KL-PPO algorithms
Researchers have demonstrated that the clipped surrogate gradient in Proximal Policy Optimization (PPO) can be precisely replicated by a Kullback-Leibler surrogate with a per-sample coefficient. This equivalence holds t…
-
New CRAX benchmark accelerates safe reinforcement learning evaluations
Researchers have introduced CRAX, a new benchmark designed to accelerate the evaluation of safe reinforcement learning (RL) agents. Built using the MuJoCo XLA physics engine, CRAX offers up to a 100x speedup compared to…