“Muscle Memory, Reaction Time, and Safety”
Keith McMillen
Sensible Robotics
August 1, 2024
Humanoid robots are inspired by human beings. Crudely at first by form but as the art advances more subtly incorporating human-like thinking and processing of events.
Presently all data collected from a robot’s sensors goes to a central AI processor. This can limit a robots’ ability to react quickly. Repetitive tasks take as much bandwidth and time as novel tasks.
As designers work to add tactile data using Sensible Robotics’ technology we hear “give me as much data as fast as you can”. This can add 10’s of MbPS to the embedded AI’s work load.
Human senses pre-process input data and send only important information gated by change. Only sending useful changing data allows faster processing and alternate communication paths.
The somatic nervous system is involved with actions that are done consciously and signals are sent from the skin and muscles to the central nervous system and vice versa.
The autonomic nervous system controls the actions that are involuntary such as the function of organs and homeostasis
Specialization of signal channels and processing has advantages of speed and reduced processor load.
Muscle memory plays a significant role in speeding up reactions by reducing the cognitive processing required for certain tasks. Muscle memory, also known as procedural memory, is a type of long-term memory that involves the learning and retention of motor skills through repetition. When a particular action is performed repeatedly, the brain creates neural pathways that allow the action to be executed more efficiently and automatically.
As per the NIH, here's how muscle memory speeds up reactions:
Automatization: Through repetition, tasks become automated, requiring less conscious effort and cognitive processing. For example, experienced musicians can play complex pieces without having to consciously think about each individual note.
Faster Execution: Muscle memory enables faster execution of learned movements because the brain doesn't need to spend as much time processing information or planning the action. This can result in quicker reaction times in situations where quick responses are required.
Efficiency in Movement: Muscle memory helps optimize movement patterns, making them more efficient. This efficiency allows for smoother and more precise movements, contributing to faster reaction times.
4. Reduced Cognitive Load: By relying on muscle memory for routine tasks, cognitive resources are freed up for other processes. This can lead to faster reactions because the brain can focus more on interpreting stimuli and making decisions rather than on executing the motor response.
Robotic network architecture must evolve to support local processing and control. Efficient operation precludes sending all sensor data to a central embedded AI that is the only source of issued control commands.
Our tactile sensors can process data at each fingertip and determine actions that can reduce reaction time and processor load.
Sensible Tactile Toolkit
Data from robotic touch sensors is complex and can be annoyingly frequent. Hi level model output clarifies gestures into essential knowledge.
In our homes using our tools, humanoid robots will need a complete understanding of how to caress, grasp and lift through:
The accumulation of grasp-related data, algorithms autonomously refine its performance by leveraging additional training data.
Detecting the onset of slippage with exceptional speed. Our system can identify slipping within milliseconds, prompting the gripper to respond quickly, potentially by applying slightly stronger force to the object.
Achieving meticulous object manipulation by applying a predetermined force. Specify your desired grasping force, and our software guarantees the object will be secured accordingly.
And in situations where your grasp isn't as stable as you'd like, we step in with expert advice. Detecting instability prompts us to offer suggestions for refining the grasp. You can opt for a stronger grip, within acceptable limits, or make precise adjustments to the grasp's position by:
Detecting the weight of an object by slightly lifting the object.
This can be used for quality control and/or for deciding the necessary grasping force.Detecting the stiffness of an object by slightly squeezing an object. This can be used, for example, to make sure not to grasp objects too hard to avoid breaking them.
Localizing objects in the hand. This function reveals the object’s relative position in relation to the gripper.
Recognizing objects from a database of previously memorized objects. By slightly touching the object, its tactile features can be recognized, which can be used to recognize objects.
This can be used, for example, to ensure that the grasped object is the correct one. It works in addition to vision, for example, when vision fails due to occlusions or poor light conditions.
Detecting shape of object based upon vector data from sensor. In addition to visual information, the local object shape can be obtained through tactile sensing. By repeatedly touching the object, the overall object shape can be obtained, and this information can be used to handle the object correctly.
Deducing the texture of an object based upon data from the multi point spatial detection and the force plate. By slightly moving across the surface of an object, its texture can be detected.
Detecting the orientation of the object within the gripper. The correct orientation of the object is crucial for various tasks, for example, when inserting the object in a hole.
And for each contact area, we will provide information about the curvature radius of an object and can recognize contact area properties such as edges, corners, and flat surfaces.