Search papers, labs, and topics across Lattice.
100 papers published across 10 labs.
Action chunk utilization triples and physical execution steps drop by over 50%, resulting in a 5.83x speedup in VLA model deployment without sacrificing performance.
Conceptual alignment in human-robot dialogue is not just a one-way street; it’s a dynamic, co-constructive process that can redefine how robots understand human intent.
Short sequences of legitimate MAVLink messages can trigger catastrophic failures in UAVs by exploiting vulnerabilities in flight-controller dynamics.
LightBenders can illuminate line drawings with surprising precision, achieving high visual quality even with notable misalignments.
The order of illuminated letters can drastically alter detection times, revealing a critical factor in visual recognition tasks.
Action chunk utilization triples and physical execution steps drop by over 50%, resulting in a 5.83x speedup in VLA model deployment without sacrificing performance.
Conceptual alignment in human-robot dialogue is not just a one-way street; it’s a dynamic, co-constructive process that can redefine how robots understand human intent.
Short sequences of legitimate MAVLink messages can trigger catastrophic failures in UAVs by exploiting vulnerabilities in flight-controller dynamics.
LightBenders can illuminate line drawings with surprising precision, achieving high visual quality even with notable misalignments.
The order of illuminated letters can drastically alter detection times, revealing a critical factor in visual recognition tasks.
Spatio-temporal reasoning boosts continuous semantic mapping accuracy by over 12%, revolutionizing how robots perceive dynamic environments.
Seamless communication between robots in challenging greenhouse environments is now achievable, thanks to a novel cloud-based architecture that integrates ROS 2 and MQTT.
Synthetic data generation via RL not only scales but also enhances generalization in bimanual dexterous manipulation by leveraging language-conditioned task annotations.
ARP not only aligns visual observations with action representations but also refines execution precision, leading to unprecedented performance in robotic manipulation tasks.
Transitioning from predictive to epistemic intelligence could revolutionize how embodied systems learn and adapt to complex environments.
The first successful sim-to-real transfer for contact-rich manipulation with tendon-driven continuum robots could redefine soft robotics applications.
Current robot memory systems fail to maintain accuracy under interference, with performance dropping sharply as unrelated sessions accumulate.
A novel hierarchical design for robotic ultrasound that enhances task success and safety by intelligently balancing high-level planning with real-time execution.
UAVs can achieve up to 94% exploration coverage by focusing on semantically rich areas, revolutionizing indoor mapping efficiency.
PenduMorph achieves unprecedented stability and reconfigurability in rolling robots, enabling new capabilities in challenging environments.
BLENDS achieves a 25.6% improvement in navigation accuracy during GNSS outages by combining Bayesian smoothing with deep learning techniques.
Filtering videos for physical consistency can boost task success rates by over 8%, bridging the simulation-to-reality gap in video generation.
A novel compliant mechanism allows rolling robots to achieve versatile motion with a simple, 3D-printable design that balances structural integrity and adaptability.
Sensor placement is the key determinant of success in dexterous manipulation, with whole-hand coverage vastly outperforming fingertip-only configurations.
FlowDPG achieves a 92% success rate in real-world robotic manipulation, bypassing the computational pitfalls of traditional policy gradient methods.
X-Safe achieves formal safety guarantees for robotic manipulation without the heavy engineering burden, enabling seamless transfer across different embodiments and tasks.
Two oblivious asynchronous robots equipped with lights can successfully form complex patterns, challenging previous assumptions about robot coordination limits.
By factoring transition extent and mode in latent actions, PoLAR achieves superior policy performance, revealing the critical role of latent action geometry in robot learning.
Achieving 6.18-9.44x faster trajectory optimization in multi-agent robotics by dynamically tuning hyperparameters at solve-time could revolutionize real-time robotic applications.
Egocentric human video can outperform traditional teleoperated robot data, achieving superior performance in embodied model pretraining with lower costs and greater diversity.
Coding agents can now autonomously refine robotic manipulation policies to achieve a staggering 99% success rate on complex tasks, revolutionizing real-world robotics.
Achieving up to 27x speedup in execution-state restoration could revolutionize low-latency AI applications, from interactive agents to robotics.
CRAX accelerates safe RL benchmarking by up to 100x, revealing critical trade-offs in performance and safety that challenge conventional wisdom in the field.
Action-aligned representations can be achieved without complex training methods, enabling robust planning in dynamic environments.
Swapping objects in 3D space can boost VLA policy success rates by over 16% on novel objects without additional data collection.
Lagrange achieves robust open-world driving by transforming decision-making into a Lagrangian action minimization problem, ensuring both interpretability and compliance with vehicle kinematics.
Advanced Vision-Language-Action models can be dramatically compressed by up to 50% without losing performance, reshaping our approach to robotic manipulation.
Continuous and consistent robotic actions can be achieved without additional network parameters, revolutionizing how robots interpret and execute complex tasks.
FlowMaps can revolutionize robotic navigation by accurately predicting object movements based on human interaction patterns, outperforming existing methods in real-world scenarios.
Achieving over 11,000× energy savings in robotic pathfinding without sacrificing decision quality could revolutionize the efficiency of mobile fulfillment systems.
Expanding training data from 2K to 238K episodes boosts dialog-driven navigation success rates by up to 100%.
UAVs can now achieve 13.82% better success rates in precise navigation tasks by effectively grounding visible targets and adapting to dynamic environments.
Reward hacking can be effectively mitigated by a novel agentic reward framework that enhances exploration in RL, leading to substantial accuracy gains in embodied world models.
Humanoid robots can now generate co-speech motions that are not only expressive but also physically executable, bridging the embodiment gap that has long hindered their performance.
Co-policy enables robots to generate musically complementary responses in real-time, transforming the landscape of human-robot musical collaboration.
EAI mobile applications are not just user interfaces; they represent a fragile cryptographic trust boundary that can be exploited to hijack physical control of AI systems.
Achieving state-of-the-art attention prediction with 99.40% lower computational costs, GazeLNN transforms how robots perceive and navigate their environments.
Bridging the gap between synthetic and real-world data could revolutionize the deployment of AI vision models in cognitive robotics.
Achieving state-of-the-art occupancy prediction while using only 2D images, Occ-VLM bridges the gap between 2D semantics and 3D understanding without the need for complex 3D inputs.
Prioritizing safety and comfort in autonomous driving can be achieved even in the face of multimodal uncertainties, thanks to a novel STL-based framework.
Post-training on synthesized safety-critical scenarios can dramatically enhance the reliability of autonomous driving systems, reducing failures in rare but critical events.
LIT-GS achieves unprecedented robustness in mapping under illumination changes by integrating LiDAR geometry, outperforming state-of-the-art methods in low-contrast environments.
A latency-resilient fusion layer transforms slow VLM outputs into real-time trajectory scoring, cutting average distance error by 30% in complex navigation scenarios.
GroundControl reveals that anticipating navigation failures through trajectory-consistent uncertainty can drastically improve the reliability of vision-language agents in real-world applications.
Large-scale human motion data can not only train robot controllers but also optimize the physical designs of robot hands, achieving superior performance in real-world applications.
Selecting the right initial noise can dramatically enhance the coherence of robot action sequences without the need for retraining or policy changes.
Achieving "collect one, get one for free" in data collection, MirrorDuo drastically enhances learning efficiency by leveraging mirrored demonstrations.
Active toes in bipedal robots can reduce energy costs by 17.5% and enhance agility by 25%, challenging previous assumptions about robotic locomotion efficiency.
Adding three in-hand degrees of freedom to a traditional gripper dramatically enhances dexterity and task feasibility in robotic manipulation.
Logic-based methods can outperform neural networks in complex assembly tasks, challenging the assumption that bigger models always deliver better results.
EquiVLA achieves a remarkable 92.6% success rate in robotic tasks by leveraging rotational equivariance, setting a new standard for generalist manipulation models.
ForEnt reveals that entrapments in forest environments can be systematically analyzed, providing critical insights for improving quadruped robot autonomy.
Preintegrating motor speeds can outperform traditional inertial measurement methods, achieving unprecedented accuracy in UAV state estimation.
Robots using the SWAP model achieved record-breaking parkour feats, including a 2.13 m jump and a 1.63 m climb, demonstrating the power of symmetry in robotic learning.
TIDY achieves unprecedented denoising performance in thermal infrared imaging, significantly enhancing robotics applications in challenging indoor environments.
Bidirectional tutoring enables robots to learn motor skills more effectively, achieving consistent behaviors and reduced reliance on guidance over time.
AUV maneuvering predictions can be dramatically improved by jointly calibrating polynomial and data-adaptive models, outperforming conventional methods in real sea-trial conditions.
Route constraints can dramatically enhance UGV localization accuracy in GNSS-degraded environments, reducing error accumulation by leveraging high-definition maps.
SurgVista achieves unprecedented visual fidelity and interaction accuracy in surgical simulations, outperforming state-of-the-art models as prediction horizons extend.
Achieving over 555 FPS in tactile simulations, TaCauchy delivers unprecedented accuracy in mechanical stress computation for robotics applications.
EventVLA's foresight-driven memory mechanism boosts long-horizon task success rates by 40% by dynamically capturing critical visual events before they vanish.
Regularizers derived from Girsanov's theorem can cut quantum control infidelity by up to 50%, transforming how we approach decoherence in open quantum systems.
A dual-agent system can autonomously translate complex biological protocols into robotic commands, bridging a critical semantic gap in laboratory automation.
Achieving safer navigation paths in unmapped environments with significantly faster computation times could revolutionize how mobile robots operate in dynamic settings.
VOiLA achieves a threefold reduction in sampling costs, making learned POMDP models viable for real-time robotic planning.
MMD-SLAM achieves unprecedented tracking accuracy and mapping quality by leveraging structural information, outperforming existing SLAM methods.
MemoryWAM achieves superior performance in robotic manipulation tasks by efficiently leveraging both short-term and long-term memory without sacrificing computational efficiency.
Genetic algorithms can significantly enhance the resilience of continuum robots by generating more diverse paths without being affected by environmental complexity.
Humanoid robot data standards could unlock the full potential of physical AI by transforming isolated datasets into a cohesive, reusable resource for robotic learning and interaction.
Temporal Self-Imitation Learning reveals that the structure of successful behaviors can serve as a powerful self-supervisory signal, drastically enhancing learning efficiency in complex tasks.
A novel continuum robot platform achieves reproducibility and ease of control, setting a new standard for research in robotic manipulation.
Robots can now localize themselves and their teammates without needing fixed infrastructure or coordinated motion, revolutionizing deployment flexibility in unstructured environments.
Achieving an 81% reduction in frequency errors during high-speed motion extrapolation could redefine the capabilities of imitation learning in dynamic environments.
3D visualization in robotic TEE reduces spatial errors by over 75%, revolutionizing operator performance and safety.
ExS2D slashes execution steps by over half while relying solely on single-arm training data, revolutionizing dual-arm manipulation efficiency.
Achieving certifiable robustness in complex control tasks, this novel MPC framework ensures stability through a unique blend of Transformer networks and Riemannian contraction analysis.
Search strategies can guarantee detection even in the face of significant perceptual uncertainties, thanks to a new detectability framework.
Explicit coordination in dual-arm robotics can boost task success rates by 27% and significantly enhance real-world performance.
A new pipeline transforms noisy sonar data into detailed 3D models of karst aquifers, overcoming significant mapping challenges.
WAMs are evolving beyond mere video generators, revealing a critical trade-off between representational richness and computational efficiency in predictive-action modeling.
Humans and VLMs show surprising consistency in their responses across diverse urban environments, challenging assumptions about geographical influence on autonomous driving performance.
A single generalist model outperforms specialized systems, achieving over 35% improvement in real-world robotic task success.
Lyapunov rewards not only improve resilience against cyber threats but also maintain low tracking errors, outperforming other reward types in critical scenarios.
Addressing domain shifts can significantly enhance the accuracy of mass estimation in rotating systems, even when the underlying physical behaviors are uncertain.
Hardware-validated simulations reveal that UAVs can achieve stable autonomous flight in complex maritime environments, bridging the gap between simulation and real-world deployment.
Robots equipped with episodic memory can boost rescue success rates by over 60% in initial interactions, transforming teamwork efficiency.
Energy-consistent URDF models reduce simulation drift by 64%, dramatically improving the realism of articulated object interactions in digital twins.
Achieving high-quality solutions for the MT-TSP-MO, our methods outperform traditional algorithms, even in challenging scenarios with numerous moving obstacles.
Iterative reinforcement learning with human feedback transforms robot gestures from stiff and unnatural to fluid and expressive, revolutionizing human-robot communication.
R2D-RL transforms the RoboCup 2D Soccer landscape, enabling seamless integration of advanced MARL techniques with a robust soccer simulation environment.
Energy-derived features can achieve competitive surface classification accuracy, even outperforming traditional inertial data in certain contexts.
EffiNav outperforms existing models in efficiency and adaptability for Object Goal Navigation, tackling the critical issue of excessive exploration in unknown environments.
NeuralMUSIC achieves superior sound source localization accuracy and robustness by seamlessly combining deep learning with classical techniques, even in challenging acoustic environments.
Human videos can now be transformed into actionable manipulation data for robots, overcoming traditional barriers in hand-object interaction estimation.
Achieving 60 FPS dynamic 4D hand reconstruction from egocentric videos, Hand-4DGS outperforms traditional methods by effectively handling occlusions and rapid motion.
