Dexterity recently expanded the deployment of its physical AI and autonomous trailer loading systems at the FedEx Hagerstown Hub in Maryland. Following an initial pilot, the company is now operating its system in a higher-volume production environment. This deployment uses Dexterity’s dual-armed Mech robot and Foresight world model to automate parcel placement within trailers.
Dexterity’s physical AI and robotic manipulation technology
Founded in 2017 by CEO Samir Menon, Dexterity, Inc. develops robotic systems for industrial logistics. Headquartered in Redwood City, California, the company focuses on material handling operations, such as truck loading, palletizing, and parcel sorting.
Its Foresight system combines visual perception, depth data, and tactile force feedback to calculate spatial trajectory and object placement decisions in real time. Foresight controls Mech, a mobile dual-armed robotic platform engineered to handle non-uniform items within tight physical boundaries like shipping trailers.
As discussed in our previous coverage of Agility Robotics and Machina Labs, robotics companies are increasingly combining robotic hardware with AI-based perception, planning, and control to handle less structured industrial operations. Dexterity’s patent portfolio reflects this approach, covering technologies related to multi-arm coordination, robotic manipulation, and automated spatial reasoning.
Exploring Dexterity’s robotics patent portfolio
Dexterity’s robotics platform combines multi-modal sensing, dynamic motion planning, and real-time control to perform complex physical tasks without relying on rigid, pre-programmed routines. By continuously processing visual, depth, and force feedback, the system coordinates manipulation and spatial positioning in changing industrial environments. The patents below detail some of the technical foundations behind these capabilities.
Handoff between human and machine
U.S. Patent No. 10,824,142 describes an architecture for maintaining operational continuity in autonomous robots through on-demand teleoperation, allowing a human operator to remotely control a robot when autonomous execution fails.
Under normal conditions, the robot uses algorithmic planning and sensor data to perform tasks such as picking, placing, and sorting. If it encounters an unmodeled obstacle or unexpected condition, such as a deformable object or unfamiliar item orientation, the system can detect that no viable execution strategy exists and automatically transition to human intervention rather than stall.
During intervention, an operator can control the robot through an input device such as a haptic interface that translates physical movements into joint torque and trajectory commands. The system can preprocess sensor-fusion data, combining visual feeds, depth maps, and overlays to support operator decisions. Meanwhile, the control framework monitors the task state and can automatically return the robot to autonomous operation once conditions become recognizable again.
The patent also describes predictive control that evaluates upcoming task steps to identify potential failures and arrange human assistance before they occur. A one-to-many architecture allows a single operator terminal to supervise multiple robots across locations, concentrating human intervention on exceptions while the broader fleet continues operating autonomously.
The patent, titled “Autonomous robot with on demand teleoperation” was filed and granted on May 1, 2018 and November 3, 2020 respectively. The listed inventors are Samir Menon, Zhouwen Sun, and Gerald Brantner, with legal representation provided by Van Pelt, Yi & James LLP.
Control system for cooperative robot handling
In high-throughput industrial settings, individual robotic arms can be limited by their weight, size, and structural constraints when handling oversized payloads.
U.S. Patent No. 12,678,959 describes a cooperative multi-robot system in which two or more autonomous manipulators jointly lift and transport heavy or unwieldy objects. The system uses an asymmetric control hierarchy, with one robot acting as the leader and others as followers. The leader determines and executes the primary motion trajectory, while followers adjust their movements relative to the leader rather than generating separate global motion plans.
Once both robots grasp the payload, the follower calculates a coordinate transformation between its end effector and the leader’s based on the payload’s dimensions. As the leader moves, it continuously shares its position and orientation, allowing the follower to update its target position and use operational space control to maintain synchronization and minimize positioning errors.
The system can also determine when a payload requires cooperative handling based on computer vision or an initial lifting failure. It then assigns robot roles, verifies both grasps, and uses force control to prevent slipping during transport. After placement, both robots release the payload and return to independent operations.
The patent entitled “Robotic system to control multiple robots to perform a task cooperatively” was filed on September 30, 2024 and granted on July 14, 2026. Zhouwen Sun, Talbot Morris-Downing, Harry Zhe Su, Rohun Kulkarni, Kevin Jose Chavez, Samir Menon, Toby Leonard Baker, David Leo Tondreau III, and Robert Holmberg are listed as the inventors. Van Pelt, Yi & James LLP provided legal representation.
Together, these patents illustrate two complementary approaches to physical AI deployment: operational resilience and physical capacity. One provides a remote human fallback for unexpected exceptions, while the other coordinates multiple robotic arms to handle oversized payloads. Combined, they help autonomous robots manage complex and unpredictable industrial workflows with greater continuity and flexibility.
Dexterity: Global Patenting Activity
Founded in 2017 out of Stanford University, Dexterity initially operated in stealth mode before publicly emerging in 2020. Backed by a Series A round and a $140 million Series B in late 2021, the company expanded its R&D and intellectual property protections. This capital influx drove priority filings to a peak in 2022, aligning with an exclusive partnership with Sumitomo Corporation. Filings normalized in 2023 after establishing core platform coverage, then rebounded in 2024 behind autonomous truck loading systems developed with FedEx and a formal joint venture with Sumitomo.
Dexterity: Top Technology Areas
Dexterity’s primary technology area is B25J, which covers robotic manipulators and end effectors. This serves as the principal mechanical framework for the company’s hardware systems. The second major classification, G05B, encompasses general control and regulating systems, protecting real-time feedback loops, sensor integration modules, and autonomous execution software. The third leading category, B65G, targets transporting and storing devices, covering conveyor integration, automated storage interfaces, and physical workflow routing in logistics operations.
The remaining portion of the portfolio covers operational software, computer vision, and industrial infrastructure. A secondary segment is dedicated to G06Q, which covers information technology for commercial processes such as inventory tracking and task management, along with B07C for automated parcel sorting. Additional lower-volume categories include B65B for packaging machinery, F16P for safety devices, and G06V alongside G06K for computer vision, image recognition, and data representation algorithms. Finally, minor filing activity in B64F covers ground installation interfaces, reflecting a patent distribution that spans physical manipulation, autonomous control, and enterprise logistics integration.
Dexterity: Top Law Firms
Dexterity distributes its patent prosecution across multiple legal practices corresponding to its primary filing jurisdictions. US-based Van Pelt, Yi & James LLP manages the largest share of applications, covering core software and hardware platforms. Taiwan-based Lee and Li, Attorneys at Law represents the second-largest volume of filings. Applications in Japan are primarily managed by Meisei International Patent Firm. The remaining portfolio is represented by European firm WP Thompson for filings before the European Patent Office, alongside additional US prosecution handled by The Marbury Law Group.
