0% found this document useful (0 votes)
2K views7 pages

Active Perception For Robot Manipulation

The document discusses active perception in robotics. It defines active perception as a robot's ability to selectively gather information about its environment through movement and exploration. This allows robots to perceive their surroundings more like animals by seeing from different viewpoints. The document reviews the history of robotics and advances that have incorporated active vision through movable cameras and sensors. It examines how active perception helps robots maintain an internal model of the world to reduce surprises and better manipulate objects while avoiding humans in complex environments.

Uploaded by

Austine Ochieng
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
2K views7 pages

Active Perception For Robot Manipulation

The document discusses active perception in robotics. It defines active perception as a robot's ability to selectively gather information about its environment through movement and exploration. This allows robots to perceive their surroundings more like animals by seeing from different viewpoints. The document reviews the history of robotics and advances that have incorporated active vision through movable cameras and sensors. It examines how active perception helps robots maintain an internal model of the world to reduce surprises and better manipulate objects while avoiding humans in complex environments.

Uploaded by

Austine Ochieng
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 7

1

A literature Review on Active perception for Robot Manipulation

Name

Course

Institution

Date
2

A literature Review on Active perception for Robot Manipulation

Introduction

Active perception can be defined as the increase of information intake from the source or flow of

data by a selection of certain behaviors that can help achieve this intended goal which is the

production of a particular environment [1,3]. This means that for a better understanding of what

is around a certain entity, there is a need to move around and explore it by sampling, observing,

feeling, and analyzing what each aspect means all based-on behavior. Active vision as a specific

point of view of active perception can be defined as a study of modeling and controlling

strategies for perception [2]. Unlike animals with a vision which can control their body and

sensory organs like eyes, body, and heads to explore their surroundings and process what to

make of it to make their decisions, the concept has been applied over the years in the field of

artificial intelligence specifically robotics to help machines perceive more of the environment by

use of cameras that are fixed on to them and can rotate also the use of partial sensors have been

used to increase the experience and make them more perceptive of the surrounding and collect

the data to make decisions on what to do next.

The main goal in these types of necessary adjustments and improvement is to ensure that like

animals, robots can see the environment from different viewpoints which can, in turn, provide a

clear and rich visual experience that will be used to learn, analyze and make a decision on it [3].

This literature review examines how active perception is applied in the robotics field and tries to

highlight the advancements made in the field to make robots behave and perceive the

environment more like animals and how it has increased productivity [2].
3

History of Robotics

Although mechanical devices have been present since even in the times of the Egyptians, the

inventions in the field of robotics saw light in the 1950s. A scientist inventor, George C. Devol

from Louisville, Kentucky made and patented a reprogrammable manipulator and named it

“Unimate” from “Universal automation” [1]. Although he did not succeed in selling the

invention for a decade, Joseph Angleberger an engineer and businessman acquired the invention

and modified it into an industrial robot which was under the company “Unimation” which was a

success and led to Engleberger being known as the father of robotics.

Over the years, research in the field increased and more inventions have been made ever since.

Academia made “Shakey” in 1958 which was an improvement of the “Unimate”. Most of the

robotic inventions at this time were able to do a limited task which was expensive with little

output, with the rising of other fields such as data science and the processing of data that comes

with it, the inspiration of sensory equipment like cameras and other sensors seemed appropriate

as it would make them more efficient in the daily working of the robots, these all began in the

late 1970s and still ongoing as with each improvement the robots can gain active vision.

Leonardo DaVinci developed drawings for a Germanic knight-shaped item that could mimic the

exact moving patterns of a man, however, it is not clear if he did make any fabrications [3,6].

Hans Bullmann, a German academic, creates humanoid androids which plays music around

1525. In 1737, French engineer Jacques Vaucanson created many artificial creatures, the far

more notable of which was a ducks that could do all the processes of a living duck except

breathing. In 1801, Joseph Jacquard, a French inventor, creates an automated loom operated by

punch cards, which marks a shift in the intended purpose of automata devices from entertaining

to industry. Babbage's theories and developments improved with his steam engines, as well as
4

the subsequent presence of George Boole's Boolean algebra, laid the groundwork for the

implementation of the newly emerging algorithmic scientific fields in the generation of different

computational models also with goal of building true nearly human skills between both the 20th

and 40th centuries. The tweentienth century is observed to have had a high impact in the

dvelopment of the history of robots and robotics. It was time when many researchers and

scientists appeared to have been so intrigued by the concept of robotics and so it rose as fast as

whirlwind. Karel Capek invented the term "robot" for the first time (what means hard work). For

his part, Isaac Asimov created the three rules of robotics in his book "Runaround," which are still

true today. Along with the emergence of related fields like as computer science and artificial

intelligence, the issue of developing automated machines/tools continues [3]. Also, the very first

controllable adjustable robotic arms was created by engineers George Devol and Joe Engleberger

in 1954. A few years after that , that would be the first robotic to be employed in industry,

specifically in the automobile industry, and the first robot to be mass-produced for these reasons.

Ever since and until today, automation has followed humankind's big events, adding to technical

advancement and new information, as well as assisting man in tough work [4,6]. Robotics were

there in the race to the moon, they are all on Mars, they have assisted in the exploration of an

ocean depths, and that they have assisted in the examination of the damages done by a nuclear

plant disaster, when a person would simply die attempting to accomplish it.

Advances in Robotics Using Active Vision

Although most parts of robotics have managed to make advances and can perform given tasks at

an improved rate than that of their predecessors, active vision is still a bigger problem. Things

like a restricted field of view, occlusion, and limited resolution all affect a robot’s ability to sense

the surrounding environment and unlike animals, this is the most important of all senses it
5

requires making its need more crucial [3,1]. In active inference, we say that the robot’s

intelligence needs to maintain a generative model where it can act according to the given

instance to reduce and avoid surprises. But due to the current level of technological

advancements, the problem of robot manipulation still presents a big challenge when working

with and around people or in highly distinct and complex environments.

With the need to ensure that robots are more of a generative model in action, some recent works

have illustrated how active vision as a form of perception emerges from active inference by

running a lot of simulations [1]. However, these approaches define the generative model in an

upfront manner which is small and discrete and not in a larger scale. This leads to the need of

robotic manipulator to aid robots in active perception. A robot manipulator is a very essential

part in robot construction as it is an electronically controlled mechanism which main goal is

interacting with the environment [3]. Some of the known manipulators known are polar,

Cartesian, cylindrical, articulated, SCARA and delta configuration which all do the main job of

manipulating the materials in surrounding environment without initiating direct physical contact

with operators and hence help in the active perception for robots.

Why Autonomous Agents Choose What to Sense in Robotics

In robotics, an autonomous agent is defined as an intelligent agent which operates on the user’s

interphases without causing any interference to the user at any instance. However, an intelligent

agent is defined as a software which performs a given set of operations on behalf of the user or

perhaps on behalf of another separate program with some level of independence by following the

user’s objectives, instructions, and goals [5,2]. These agents may include systems in a

technological or surrounding habitat which perceives all or even some state of that surroundings

then reacts around it to forward its very own goal is an example of an agents. Drivers shape one
6

such objective. The agency modifies what it perceived by acting to modify an element of the

surroundings or its condition.

The activity, or lack thereof, is the essential distinction with an active vision technology and

other perceiving methods. While both structures have choice elements, just the proactive method

combines both outside (through motors) and interior (by system parameters) dynamically

modulation formats to the total agent's behavior [4]. Imagine the following scenario: we

successfully developed filters to detect a certain object, such as a sword, from photos using

cutting-edge machine learning approach. Take into account the fact that our filter has a 90%

successfulness. That might be a ground-breaking conclusion, but it may not be as exciting for a

robotic system that is functioning [2]. The core of proactive perceiving is to select a goal

predicated on a present world view and then put into motion the behaviors that will help you

attain it. In those other terms, proactive perceiving is purposeful because it must synthesize

information and movement in order to attain a goal [6]. To build a concept, one must have a

foundation for understanding the set of objectives, such as behavioral arithmetic.

Conclusion

From the history of robotics, it is clear that robotics has taken huge strands and even more in the

current modern world. There are still many more sophisticated innovations that the world is yet

to witness. However, we have also observed that robots still need a lot of work to be done on

them. This will ensure that like animals, robots can see the environment [4] from different

viewpoints which can, in turn, provide a clear and rich visual experience that will be used to

learn, analyze and make a decision on it.


7

Bibliography

[1]

“History of robotics,” BBVA.CH. https://www.bbva.ch/en/news/history-of-robotics/ (accessed

May 11, 2022).

[2]

R. Bajcsy, Y. Aloimonos, and J. K. Tsotsos, “Revisiting active perception,” Autonomous Robots,

vol. 42, no. 2, pp. 177–196, Feb. 2017, doi: 10.1007/s10514-017-9615-3.

[3]

C. Premebida, R. Ambrus, and Z.-C. Marton, Intelligent Robotic Perception Systems.

IntechOpen, 2018. [Online]. Available: https://www.intechopen.com/chapters/62978

[4]

K. Greene, “MIIS2: The Perception Process,” introtocommopensource.ridgewater.edu, Jul. 07,

2018. http://introtocommopensource.ridgewater.edu/ModuleII/ModIISect2.html

[5]

Q. Le, A. Saxena, and A. Ng, “Active Perception: Interactive Manipulation for Improving Object

Detection.” Accessed: May 11, 2022. [Online]. Available:

http://ai.stanford.edu/~quocle/activevision.pdf

[6]

“APRIL,” Advanced Robotics. https://advr.iit.it/index.php/research/april (accessed May 11,

2022).

You might also like