Franka Research 3
About
About Franka Research 3
Franka Research 3 is a seven-axis, torque-sensitive robot arm for manipulation, control and physical-AI research. It provides an 855 mm reach, 3 kg payload, joint torque sensing and low-level 1 kHz access through the Franka Control Interface.
Developers can use libfranka, ROS 2, MATLAB or Simulink integrations and the robot model library for custom control, sensing and data collection. The system is designed for laboratories and research development. Production deployment and operator safety require the appropriate Franka product, safeguards and application engineering.
Capabilities
Source-backed capabilities
- Capability areas: Collaborative operation, Torque-controlled robot arm
- Workflows: Robot manipulation research
- Force-controlled manipulation
Claimed - Franka Robotics presents Franka Research 3 as a force-sensitive research platform with joint torque sensing and external control interfaces for controlled manipulation experiments; the brochure is a product claim, not deployment evidence.
Source - Object manipulation
Claimed - Franka Robotics presents Franka Research 3 as a force-sensitive research platform with joint torque sensing and external control interfaces for controlled manipulation experiments; the brochure is a product claim, not deployment evidence.
Source - Remote and shared control
Demonstrated - 2025-11-19 - Legacy MANUS Quantum Metagloves and Meta Quest 3 supplied the teleoperation inputs.
Source
Observed use
Exact-product configurations are shown without inferring supply, purchase, customer status, standard fit, deployment or partnership.
- MANUS Quantum Metagloves
2025-11-19 - Research Experiment - Controlled By - Physical - ByteDexter research used legacy MANUS Quantum Metagloves and Meta Quest 3 to teleoperate a 20-DoF ByteDexter hand on a Franka Research 3 arm.
Source - Observed configuration
Research Experiment - Tested With - Physical - ORD-WM experiments use a Franka Research 3 with three calibrated RealSense D435 cameras for object-state observations.
Source - Two external cameras and one wrist camera; exact models unspecified
Research Experiment - Connected - Physical - LIBERO-Dynamic physical experiments use Franka Research3 with two external camera views and one wrist camera.
Source - Sharpa Wave
Digital Simulation - Mounted - Simulated - BiDex-100 simulates two Franka Research3 arms with22DoF Sharpa Wave hands for96tasks.
Source
Research
- Self-Distilling Future Video Representations for World Action Models
2026-10-21 - Research Publication - FRED improves world action models by self-distilling future video representations and conditioning them on observation history. The KAIST and RLWRLD team evaluates robot manipulation on Franka Research 3 and OpenArm hardware, including transfer from human demonstration videos. Author-reported success on tasks absent from robot training data reaches 50.0%, compared with 19.4% for the strongest evaluated WAM baseline.
Source - PaceVLA: Adaptive Temporal Action Quantization in Vision-Language-Action Models
2026-10-19 - Research Publication - PaceVLA lets a vision-language-action policy select action granularity and prediction horizon during execution. Three simulation benchmarks and six physical tasks test faster manipulation across multiple robot embodiments.
Source - Embodied Playground: A Long-Horizon Agentic Robotics Benchmark for Physical, Strategic, and Spatial Reasoning
2026-10-19 - Research Publication - Embodied Playground tests long-horizon robotic reasoning across 410 simulated task environments. Its EP-100 subset contains 100 tasks in five suites and 16 task types, with separate goal, rule-compliance and physical-validity checks. The research team includes Sharpa and university affiliations.
Source - BiDex-100: A Comprehensive Benchmark for Bimanual Dexterous Manipulation Across 100+ Simulation Tasks
2026-10-18 - Research Publication - BiDex-100 is a simulation benchmark with 106 bimanual dexterous tasks across five capability axes. It supplies8,080 validated teleoperated demonstrations across dual-Franka and Astribot S1 simulation embodiments.
Source - CusADi2: Latency-Optimized Batch Computation of Robot Kinematics, Jacobians, and Hessians on the GPU
2026-10-18 - Research Publication - CusADi2 reduces batch-computation latency by parallelizing both across and within robot-model evaluations on GPUs. The authors benchmark kinematics, Jacobians and Hessians across67 MuJoCo Menagerie robots and report lower latency at batch sizes64–4096 than evaluated alternatives.
Source - LIBERO-Dynamic: Benchmarking Robotic Manipulation in Dynamic Environments
2026-10-17 - Research Publication - LIBERO-Dynamic extends forty manipulation tasks to moving targets while accounting for policy inference latency. Simulation experiments compare six policies, supplemented by a small physical study using Franka Research 3.
Source - ORD-WM: Learning Object-Centric Rigid Dynamics for 3D World Modeling in Robotic Manipulation
2026-10-16 - Research Publication - ORD-WM learns object-level rigid transformations in a geometry-aware 3D latent space, separating persistent object shape from pose changes. Simulated and Franka Research 3 experiments examine multi-step prediction and closed-loop manipulation planning.
Source - YUBI-Real2Sim: 3DGS Scene Verification for Reliable Policy Evaluation
2026-10-13 - Research Publication - YUBI-Real 2 Sim studies whether Gaussian-splatting scene reconstructions preserve robot-policy rankings under moving fisheye wrist cameras. It reproduces real ordering in eight of nine scene–task comparisons, then tests how background artifacts and capture coverage affect evaluation reliability.
Source - Beyond the End-Effector: Learning Whole-Arm Contact Manipulation from Human Demonstrations
2026-10-13 - Research Publication - WACL transfers human demonstrations into whole-arm contact manipulation, including tasks with constrained end-effector objectives. Physical evaluations use UFACTORY xArm 7 and zero-shot transfer to Franka Research 3.
Source - StreamingWAM: Streaming World Action Modeling via Chunk-Wise Staircase Flow Matching
2026-10-11 - Research Publication - StreamingWAM uses chunk-wise staircase flow matching, prefix rollout and video-history memory to produce smoother continuous robot control. Author-reported tests span LIBERO, LIBERO-Plus, RoboTwin 2.0 and physical Franka Research 3 manipulation; code, a LIBERO checkpoint and real-world demonstration data are linked.
Source - Dynin-Robotics: Omnimodal Unified Diffusion Vision-Language-Action Model
2026-09-11 - Research Publication - Dynin-Robotics uses a shared masked-diffusion model for robot policy generation, future visual states, goal prediction and instruction reconstruction. It reports 78.4% average success across four manipulation conditions on Franka Research 3, 98.1% on LIBERO and 73.0% on zero-shot LIBERO-Plus. An optimised decoder achieves up to 29.2× model-side action-decoding speedup under the study’s profiling setup. Code and model release are planned.
Source - Axis Sim Dataset V1
2026-09-04 - Research Publication - Axis Sim Dataset V1 provides diverse simulation demonstrations, training code and reproducible benchmarks for continual pretraining of robot policies.
Source - Robust Operational Space Control with Conformal Disturbance Bounds for Safe Redundant Manipulation
2026-07-01 - Research Publication - Redundant robotic manipulators operating in constrained and human-interactive environments require accurate task-space tracking together with rigorous safety guarantees under dynamic uncertainties. Classical operational space computed torque controller (OSCTC) relies on accurate dynamic models and degrades in the presence of disturbances. In contrast, the data-driven paradigm of residual learning approximates disturb
Source - Sequential Planning via Anchored Robotic Keypoints
2026-06-29 - Research Publication - We present Sequential Planning via Anchored Robotic Keypoints, SPARK, a training-free neurosymbolic manipulation system that reaches 43.7% on six LIBERO-PRO position \& task cells, more than doubling CaP-Agent0 and Vision-Language-Action (VLA) baselines. CaP-Agent0, a multi-turn code-generation agent, achieves 18.2% by re-querying an LLM at every turn, but its restart-from-scratch solution proves costly against m
Source - WatchAct: A Benchmark for Behavior-Grounded Robot Manipulation
2026-06-24 - Research Publication - A robot working alongside people must reason about what they have done, in what order, and with what intent. Video carries the spatial layouts, object histories, and gestures that language leaves underspecified, yet today's manipulation benchmarks pair an instruction with a single current image, offering no way to evaluate reasoning over observed human behavior. We introduce WatchAct, a benchmark for robot manipulati
Source - Backpropagating Through Simulation: Analytic Policy Gradients for Sample and Learning Efficient Differentiable Continuous Control
2026-06-19 - Research Publication - Model-free reinforcement learning algorithms such as Proximal Policy Optimization (PPO) treat the environment as a black box, estimating policy gradients from sampled rewards; this process demands millions of interactions and relies on high-variance advantage estimates. When environment dynamics are differentiable, the return is an end-to-end differentiable function of the policy parameters, enabling exact gradient c
Source - Object-Centric Residual RL for Zero-Shot Sim-to-Real VLA Enhancement
2026-06-17 - Research Publication - Vision-Language-Action (VLA) models can generalize across diverse manipulation tasks, but their imitation-learning-based policies remain brittle in precise physical interactions due to compounding execution errors; Can a reinforcement learning policy trained purely in simulation improve the robustness of real-world VLAs zero-shot? Residual RL, which learns a corrective policy on top of a frozen VLA, offers a natural
Source - Real-World Deployment of Massively Parallel Sampling-Based MPC for Contact-Rich Manipulation
2026-06-16 - Research Publication - Sampling-based Model Predictive Control (SMPC) is a promising strategy for contact-rich robotic manipulation, combining gradient-free optimization with massively parallel GPU simulation. Yet, most prior work relies on simplified dynamics or remains confined to simulation. We present an MPC framework that leverages JAX for large-scale parallelization and efficient computation, coupled with the high-fidelity MuJoCo MJX
Source - WangLei-ZJU-CSC/pi05-fr3-reproduction
2026-05-28 - Research Publication - Local reproduction of π0.5 on real Franka Research 3 robots with VR teleoperation and LoRA fine-tuning.
Source - PhAIL: A Real-Robot VLA Benchmark and Distributional Methodology
2026-05-28 - Research Publication - Launches a real-robot VLA benchmark on Franka FR3 with distributional metrics beyond binary success rates.
Source - Rethinking Muon Beyond Pretraining: Spectral Failures and High-Pass Remedies for VLA and RLVR
2026-05-19 - Research Publication - Muon is a matrix-aware optimizer that leverages Newton-Schulz (NS) iterations to enforce spectral gradient orthogonalization by driving all singular values of the momentum matrix toward 1. While this uniform spectral whitening enhances exploration and outperforms AdamW in LLM pretraining, we show it could lead to fundamental limitations beyond pretraining in two regimes: (i) cross-modality vision-language-action (VLA
Source - XingxinHE/franka_moveit_camera_calibration
2026-05-05 - Research Publication - RealSense camera hand-eye calibration with a Franka Research 3 robot using ROS 2 and MoveIt. No Docker required.
Source - EmbodiedForge/droid-fr3
2026-04-29 - Research Publication - DROID robot platform adapted for Franka Research 3 (FR3) - replaces polymetis with Franky SDK
Source - Passive iFIR filters for data-driven velocity control in robotics
2026-03-31 - Research Publication - We present a passive, data-driven velocity control method for nonlinear robotic manipulators that achieves better tracking performance than optimized PID with comparable design complexity. Using only three minutes of probing data, a VRFT-based design identifies passive iFIR controllers that (i) preserve closed-loop stability via passivity constraints and (ii) outperform a VRFT-tuned PID baseline on the Franka Researc
Source - Form-Fitting, Large-Area Sensor Mounting for Obstacle Detection
2026-03-24 - Research Publication - We introduce a low-cost method for mounting sensors onto robot links for large-area sensing coverage that does not require the sensor's positions or orientations to be calibrated before use. Using computer aided design (CAD), a robot skin covering, or skin unit, can be procedurally generated to fit around a nondevelopable surface, a 3D surface that cannot be flattened into a 2D plane without distortion, of a robot. T
Source - Learning Adaptive Force Control for Contact-Rich Sample Scraping with Heterogeneous Materials
2026-03-11 - Research Publication - Their work proposes an adaptive control framework to address this, relying on a low-level Cartesian impedance controller for stable and compliant physical interaction and a high-level reinforcement learning agent that learns to dynamically adjust interaction forces at the end-effector.
Source - Learning-Based Robust Control: Unifying Exploration and Distributional Robustness for Reliable Robotics via Free Energy
2026-03-06 - Research Publication - Inspired by the free energy principle in computational neuroscience, to address these challenges, the authors propose a model for policy computation that jointly learns environment dynamics and rewards, while ensuring robustness to epistemic uncertainties.
Source - GaussTwin: Unified Simulation and Correction with Gaussian Splatting for Robotic Digital Twins
2026-03-05 - Research Publication - Digital twins promise to enhance robotic manipulation by maintaining a consistent link between real-world perception and simulation. However, most existing systems struggle with the lack of a unified model, complex dynamic interactions, and the real-to-sim gap, which limits downstream applications such as model predictive control. Thus, we propose GaussTwin, a real-time digital twin that combines position-based dynam
Source - Design, Mapping, and Contact Anticipation with 3D-printed Whole-Body Tactile and Proximity Sensors
2026-03-05 - Research Publication - Robots operating in dynamic and shared environments benefit from anticipating contact before it occurs. We present GenTact-Prox, a fully 3D-printed artificial skin that integrates tactile and proximity sensing for contact detection and anticipation. The artificial skin platform is modular in design, procedurally generated to fit any robot morphology, and can cover the whole body of a robot. The skin achieved detection ranges of up to 18 cm during evaluation. To characterize how robots perceive nearby space through this skin, we introduce a data-driven framework for mapping the Perisensory Space -- the body-centric volume of space around the robot where sensors provide actionable information for contact anticipation. We demonstrate this approach on a Franka Research 3 robot equipped with five GenTact-Prox units, enabling online object-aware operation and contact prediction.
Source - VLAD-Grasp: Zero-shot Grasp Detection via Vision-Language Models
2025-11-08 - Research Publication - Towards this, the authors present VLAD-Grasp, a Vision-Language model Assisted zero-shot approach for Detecting Grasps.
Source - Dual-Stream Diffusion for World-Model Augmented Vision-Language-Action Model
2025-10-31 - Research Publication - To address this, the authors propose DUal-STream diffusion (DUST), a world-model augmented VLA framework featuring a multimodal diffusion transformer that maintains separate modality streams while enabling cross-modal knowledge sharing.
Source - How Well Do Latent World Models Understand Partially Observable Safety Constraints?
2025-10-07 - Research Publication - The authors introduce two diagnostics for these gaps: a mutual-information-based measure of safety observability and a rollout-based measure of future safety predictability.
Source - Gaussian Process-Based Active Exploration Strategies in Vision and Touch
2025-07-07 - Research Publication - This work proposes fusing visual and tactile observations into a unified Gaussian Process Distance Field (GPDF) representation for active perception of object properties.
Source - Proactive Hierarchical Control Barrier Function-Based Safety Prioritization in Close Human-Robot Interaction Scenarios
2025-05-21 - Research Publication - This paper presents a hierarchical control framework based on Control Barrier Functions (CBFs) designed to ensure safe and adaptive operation of autonomous robotic manipulators during close-proximity human-robot interaction.
Source - Safe Multi-Robotic Arm Interaction via 3D Convex Shapes
2025-03-14 - Research Publication - The authors present an online collision avoidance methodology leveraging High-Order Control Barrier Functions (HOCBFs) constructed for safe interactions among 3D convex shapes to address this issue.
Source - Generalizing Safety Beyond Collision-Avoidance via Latent-Space Reachability Analysis
2025-02-02 - Research Publication - The authors propose Latent Safety Filters, a latent-space generalization of HJ reachability that tractably operates directly on raw observation data (e.g., RGB images) to automatically compute safety-preserving actions without explicit recovery demonstrations by performing safety analysis in the latent embedding space of a generative world model.
Source - GenTact Toolbox: A Computational Design Pipeline to Procedurally Generate Context-Driven 3D Printed Whole-Body Artificial Skins
2024-12-01 - Research Publication - Developing whole-body tactile skins for robots remains a challenging task, as existing solutions often prioritize modular, one-size-fits-all designs, which, while versatile, fail to account for the robot's specific shape and the unique demands of its operational context.
Source - Goal-oriented Semantic Communication for Robot Arm Reconstruction in Digital Twin: Feature and Temporal Selections
2024-11-13 - Research Publication - To this end, the authors propose a novel goal-oriented semantic communication (GSC) framework to extract the GSC information for the robot arm reconstruction task in the DT, with the aim of minimising the communication load under the strict and relaxed reconstruction error constraints.
Source - Collision Avoidance for Convex Primitives via Differentiable Optimization Based High-Order Control Barrier Functions
2024-10-24 - Research Publication - This work addresses these challenges by introducing a high-order CBF (HOCBF) framework for collision avoidance among convex primitives.
Source - Learning Tactile Insertion in the Real World
2024-05-01 - Research Publication - The study proposes to use Reinforcement Learning to learn an end-to-end policy, mapping directly from tactile sensor readings to actions.
Source - Constrained Passive Interaction Control: Leveraging Passivity and Safety for Robot Manipulators
2024-03-14 - Research Publication - Passivity is necessary for robots to fluidly collaborate and interact with humans physically. Nevertheless, due to the unconstrained nature of passivity-based impedance control laws, the robot is vulnerable to infeasible and unsafe configurations upon physical perturbations. In this paper, we propose a novel control architecture that allows a torque-controlled robot to guarantee safety constraints such as kinematic limits, self-collisions, external collisions and singularities and is passive only when feasible. This is achieved by constraining a dynamical system based impedance control law with a relaxed hierarchical control barrier function quadratic program subject to multiple concurrent, possibly contradicting, constraints. Joint space constraints are formulated from efficient data-driven self- and external C^2 collision boundary functions. We theoretically prove constraint satisfacti
Source - Grounding LLMs For Robot Task Planning Using Closed-loop State Feedback
2024-02-13 - Research Publication - The study introduces a novel approach to task planning that utilizes two separate LLMs for high-level planning and low-level control, improving task-related success rates and goal condition recall.
Source - Gavin1129/franka_haptic_teleoperate
2023-12-18 - Research Publication - Franka research 3 robot arm teleoperated by sigma 7 Haptic device
Source - Domain Randomization for Sim2real Transfer of Automatically Generated Grasping Datasets
2023-10-06 - Research Publication - To avoid constraining the operational space, an increasing number of works propose grasping datasets to learn from.
Source - Toward a Plug-and-Play Vision-Based Grasping Module for Robotics
2023-10-06 - Research Publication - This paper introduces a vision-based grasping framework that can easily be transferred across multiple manipulators.
Source - Differentiable Optimization Based Time-Varying Control Barrier Functions for Dynamic Obstacle Avoidance
2023-09-29 - Research Publication - Control barrier functions (CBFs) provide a simple yet effective way for safe control synthesis. Recently, work has been done using differentiable optimization (diffOpt) based methods to systematically construct CBFs for static obstacle avoidance tasks between geometric shapes. In this work, we extend the application of diffOpt CBFs to perform dynamic obstacle avoidance tasks. We show that by using the time-varying CBF (TVCBF) formulation, we can perform obstacle avoidance for dynamic geometric obstacles. Additionally, we show how to extend the TVCBF constraint to consider measurement noise and actuation limits. To demonstrate the efficacy of our proposed approach, we first compare its performance with a model predictive control based method and a circular CBF based method on a simulated dynamic obstacle avoidance task. Then, we demonstrate the performance of our proposed approach in expe
Source - BolunDai0216/fr3_ros
2022-11-17 - Research Publication - ROS package for Franka Research 3 robot.
Source
Resources
Media library
- Franka Research 3 system view
image · Franka Robotics
Source - Franka Research 3 joint detail
image · Franka Robotics
Source - Franka Research 3 in a research lab
image · Franka Robotics
Source - Franka Research 3 reference platform
video · Franka Robotics
Source
Official Franka Robotics product references, current manuals, developer documentation and exact product media for the Franka Research 3 robot system.
- Franka Research 3 product reference
WEB · Franka Research 3 · Franka Robotics
Source - Franka Research 3 brochure
PDF · Franka Research 3 · Franka Robotics
Source - Franka Research 3 hardware manual
PDF · FR3 arm v2.1 and v2.2 · Franka Robotics
Source - Franka Research 3 operating manual
PDF · System Image 5.10 · Franka Robotics
Source - Franka Research 3 datasheet
PDF · Franka Research 3 · Franka Robotics
Source - Franka Control Interface documentation
WEB · Franka Research 3 and Franka Control Interface · Franka Robotics
Source
Timeline
- ByteDexter demonstrates 20-DoF hand teleoperation
2025-11-19 - Demonstration - The physical setup mounted a ByteDexter hand on a Franka Research 3 arm. Legacy MANUS Quantum Metagloves and Meta Quest 3 supplied the teleoperation inputs.
Source