How Robots Figure Out What They're Carrying While Moving

This patent describes how a mobile robot can estimate the weight and balance of an object it's carrying *while moving it*, and then adjust its movements for better control and safety.

What it covers

This patent details a method for a mobile robot to dynamically understand the properties of an object it is manipulating. First, the robot plans an initial path, or "trajectory," for the object (Claim 28). As the robot begins to move the object along this path, it simultaneously estimates the object's "mass characteristics" — things like its total weight, where its center of gravity is, and how hard it is to rotate (Claim 36). It does this by measuring the object's acceleration and the forces and torques, or "wrench," applied to it (Abstract). Based on these real-time estimates, the robot then modifies its operation, which can involve adjusting the motion of its robotic arm (Claim 32), changing the torques at its arm joints (Claim 33), or even adjusting the movement of its mobile base (Claim 34). For example, a robot might pick up a box, start moving, realize the box is heavier than expected, and then automatically slow down or adjust its posture to prevent tipping.

What it doesn't cover

• —Does not cover estimating an object's mass characteristics *before* the object is in motion, such as by placing it on a scale.
• —Does not cover robots that estimate mass characteristics but *do not* then modify their subsequent operation based on those estimates.
• —Does not cover robots that are stationary and only manipulate an object with an arm without a mobile base.
• —Does not cover estimating mass characteristics without using both determined accelerations and sensed wrenches.
• —Does not cover modifying robot operation without planning a *new* trajectory or adjusting arm/base motion based on the estimated mass characteristics.

The clever bit

The core innovation is the ability to estimate an object's mass properties *while it's actively being moved* and then immediately use that information to adjust the robot's actions. This dynamic feedback loop allows robots to adapt on the fly, rather than relying on pre-programmed assumptions about what they are carrying.

Why it matters

This technology is crucial for making mobile robots more adaptable and safer when operating in unpredictable environments or handling objects of unknown properties. By understanding an object's mass in real-time, robots can avoid dropping items, prevent damage to themselves or their surroundings, and perform tasks more efficiently. This capability is especially important for robots designed to work alongside humans or in dynamic settings like warehouses and construction sites.

Real-world examples

• 1.Boston Dynamics' Atlas humanoid robot
• 2.Boston Dynamics' Spot robot with a robotic arm
• 3.Warehouse logistics robots handling varied packages
• 4.Construction robots manipulating building materials
• 5.Robots assisting in disaster relief, moving unknown debris