Boston Dynamics Rebuilds Stretch Gen 2 to Unload 1,000 Boxes per Hour

Boston Dynamics has introduced the second generation of Stretch , its mobile robot for unloading floor-loaded trailers and shipping containers, where workers stack cartons directly on the floor. The redesign targets the failures that interrupt dock operations: loose connectors, worn wiring, damaged panels and drive modules that take hours to replace. Reducing those interruptions can raise effective throughput and limit service calls.

Multipick reaches the peak rate by grasping multiple compatible cases during one cycle. The release provides no sustained-rate distribution by carton size, stack quality or trailer condition. Downstream conveyor capacity and the share of cases eligible for multipick will also affect completed cases per hour.

100,000 hours reshaped the hardware

Boston Dynamics used data from more than 100,000 operating hours and tens of millions of unloaded boxes to identify common causes of unplanned downtime. The team cut internal cabling by 81%, removing about half a mile of wire and reducing opportunities for intermittent faults, abrasion and connector failures.

• Connectors. Heavy-duty, automotive-grade connectors withstand vibration from rapid box handling and travel over uneven floors.
• Gripper. Revised valve systems and more precise sensing support faster, more reliable picks.
• Joints. Continuous-rotation joints spin without striking hard stops that previously concentrated mechanical stress.
• Exterior controls. A lower, recessed emergency-stop button reduces accidental activation by falling boxes, while revised panel materials resist impact damage.

Maintenance received the same attention. A swerve-drive module combines wheel propulsion and steering, allowing the mobile base to change direction in confined spaces. Replacing a complete module now takes about one hour, down from a full day. The redesign also removes flexing dynamic cables and complex harnesses, standardizes high-strength fasteners to reduce tool changes, and adds diagnostics that help customer technicians complete routine repairs.

Fleet data sharpens the software

Stretch Gen 2 runs on a redesigned onboard computing platform that supports fleet-wide software releases without hardware changes. Boston Dynamics says those updates improve case detection, motion planning, pick selection and world modeling. In practical terms, the software identifies cartons, chooses collision-free arm paths, selects grasp points and maintains an internal representation of the trailer and surrounding equipment.

Tens of millions of completed picks give the perception system examples of package and stack variation that are difficult to reproduce cheaply in simulation. According to the company, Stretch arrives pretrained to detect most packages without site-specific setup. That installed base can also expose rare failure cases for later software releases.

Purpose-built for the trailer

Trailer unloading combines repetitive lifting, confined workspaces, variable temperatures and irregular stacks. Boston Dynamics calls Stretch the market leader and claims more deployments than any competing logistics robot, although the announcement gives no total deployment count. Early customers include DHL Supply Chain, Gap, H&M and Maersk’s Performance Team.

Stretch uses purpose-built hardware for a defined warehouse workflow. Boston Dynamics is also developing Atlas, while Figure and Agility are pursuing humanoid systems for broader industrial work. Stretch’s narrower scope concentrates engineering, training data and maintenance procedures around one repeatable task.

Boston Dynamics launched Gen 2 during a leadership transition, having recently named Rohit Prasad CEO. Prasad previously led Amazon’s Alexa artificial general intelligence effort. He inherits a logistics product moving from early deployments toward longer-running fleets, where repair time, spare-parts access and predictable availability shape operating costs.

The deployment math remains open

Boston Dynamics’ product page emphasizes uptime and serviceability. Public materials leave several procurement and integration details unspecified:

• Performance: Sustained throughput by carton mix, trailer condition and multipick eligibility.
• Reliability: Fleet availability, mean time between failures and mean time to repair.
• Power: Shift duration, duty cycle, battery degradation and charging cadence behind the two-shift claim.
• Integration: Interfaces for warehouse management systems, conveyors and safety programmable logic controllers, plus network, security and software rollback requirements.
• Service: Diagnostic access, spare-parts lead times, technician training, warranty coverage and remote-support terms.
• Rollout: Pricing, shipment dates, supported regions and an upgrade path for first-generation fleets.

Effective throughput combines pick speed with availability, charging time, recovery time and downstream flow. Acceptance testing can measure completed cases per scheduled hour, interventions per trailer, recovery duration and the share of repairs completed by warehouse technicians. Those figures will show whether Gen 2’s mechanical changes translate into lower support costs and more predictable fleet operations.