The Revolution in Human-Centric Robots:
Rod Brooks, the former CTO of iRobot and current CTO of Rethink Robotics, presented a new class of industrial robots called Baxter. Baxter is designed to be easier to install and program, allowing small businesses to incorporate robots without the need for engineers.

The Two Themes Driving Baxter’s Development:
Human-centric robots: Baxter is designed to be safe and easy for ordinary people to use, breaking the barrier between humans and robots in industrial settings. Rethinking manufacturing: Baxter aims to address the challenges of manufacturing consumer goods in the United States.

The Rise of Ordinary People Touching Computers:
In the past, computers were hidden away from ordinary people due to their sensitivity to smoke particles and the need for specialized knowledge to operate them. The introduction of the Apple IIe and the PC in the 1980s brought computation to ordinary people, empowering them with access to and control over information.

Extending the Empowerment to Robots:
Robots, especially industrial robots, are typically kept away from ordinary people due to safety concerns. However, the last 10 years have seen an increase in robots that ordinary people can touch, such as surgical robots, warehouse robots, and home cleaning robots.

Factors Contributing to the Uptake of Robots:
The exponential decrease in the cost of computation and sensors over the last 50 years has made robots more accessible. Non-robotic applications have driven down the cost of sensors, making them more affordable for robotic applications.

Baxter’s Public Debut:
Baxter’s first public appearance will be at Carnegie Mellon University on Monday, followed by a public showcase in Pittsburgh the following Monday.

Consumer Goods and Robotics:
The cost of sensors and components for robots has decreased significantly due to consumer electronics, making robot construction more affordable.

Advancements in Computer Vision and SLAM:
Improvements in computer vision and simultaneous localization and mapping (SLAM) have enhanced robots’ effectiveness in navigation and understanding their surroundings.

User-friendly Robotics:
Certain tasks have reached a usability threshold where ordinary individuals can interact with robots without specialized training.

Disruptive Price Point:
The Roomba’s success can be attributed to its disruptive price point of $200, making it an impulse purchase rather than a luxury item.

Lesson 1: Disruptions Create New Markets:
Disruptive technologies can create new markets by offering affordable solutions to problems that people didn’t realize they had.

Lesson 2: Real Robots Have Real Users:
Engineers and technologists are not representative of the end-users of robots. Real users only care about getting their tasks done, not the coolness or complexity of the technology.

Simplicity in Interaction:
Over time, the Roombas evolved to have simpler user interfaces, reducing support calls about basic operations like turning the device on or off.

User-Centric Design:
Rodney Brooks emphasizes the importance of user-centric design in robotics. The Roomba, a robotic vacuum cleaner, was designed with a simple, one-button interface to make it easy for users to operate. Users often find unconventional ways to use robots, such as stacking virtual walls on top of the Roomba to create Breitenberg Vehicles that avoid each other and spread out to clean a floor.

User-Friendly Interfaces for Military Robots:
In Iraq and Afghanistan, military robots were used to deal with improvised explosive devices. The iRobot and Foster Miller robots were two major types used, but the iRobot’s complex user interface was criticized by users. A game controller interface was introduced, and with just five minutes of practice, users could effectively operate the robot, achieving 85% of its capabilities.

Low-Cost Manufacturing in China:
Rodney Brooks visited China to explore low-cost manufacturing for robot toys and Roombas. Chinese factories assemble robots by hand, using simple infrastructure and tools. Despite technological advancements, consumer products like iPads and Roombas are still largely assembled by hand.

Key Trends in Manufacturing:
Since World War II, manufacturing has been outsourced to low-cost labor regions, starting with Japan, then South Korea, Taiwan, China, and now Thailand, Indonesia, and Malaysia. The cost of manufacturing in China has increased rapidly in recent years due to rising wages and labor unrest. There is a growing demand for responsive and short supply chains, as well as manufacturing close to innovation centers. Intellectual property concerns and rising oil costs are also driving companies to bring manufacturing back to the US.

Challenges with Current Industrial Robots:
Traditional industrial robots are expensive, complex to use, and require extensive integration costs. They are not safe for people to be around and lack adaptability and flexibility. They are not suitable for small production runs and are inaccessible to many small manufacturing companies.

Rethink Robotics’ Approach:
Rethink Robotics aims to make robots easy to use, low-cost, and suitable for small production runs. The company’s first product, Baxter, is a robot designed to work alongside people in factories. Baxter is easy to program and can be deployed quickly without extensive integration costs.

Baxter’s Features and Benefits:
Baxter has a friendly and approachable design, with rounded and square heads and orange and red colors. It is easy to program, with a simple user interface that does not require specialized knowledge like quaternions. Baxter is safe to work with, with built-in sensors and software that prevent it from harming people. It is versatile and can perform a wide range of tasks, including assembly, packaging, and quality control.

Conclusion:
Rethink Robotics believes that Baxter and similar robots can disrupt the manufacturing industry by making robots accessible to small and medium-sized companies and enabling them to compete with larger manufacturers. The company’s goal is to bring manufacturing back to the US and create new jobs.

Features of Baxter:
Behavior-based software comes preloaded, no programming required. Two 7-degree-of-freedom arms with series elastic actuators and force sensing. Cameras in each wrist and chest, and one above the face. Compliant arms for safe operation near humans. Five-pound payload and 6,300 hours of continuous operation at maximum acceleration. Interchangeable end effectors for various tasks. Non-mobile base with optional casters for easy relocation. Simple training process by guiding the arms and using the graphical user interface on the screen. Adapts to changes in the environment and can perform a wide range of simple tasks. Awareness of its job, environment, people, conveyables, boxes, and certain factory categories.

Benefits of Baxter:
Easy to use, no systems integration or programming required. Fast training time, tasks can be changed quickly. Extensible platform with new software and apps to increase capabilities. Series elastic actuators provide natural springiness and adjustable compliance. Gravity compensated zero force mode allows easy movement of the arms. Simple programming methodology for ordinary factory workers. Affordable price of $22,000 for the main robot. Distributed supply chain in the USA and Canada for efficient manufacturing.

How Baxter Works:
To train Baxter, grab its arm, and it will go into gravity compensated zero force mode. Use the buttons on the cuff to give commands and interact with the graphical user interface. The robot infers primitive operations from the user’s actions, such as pick up and put down. Safety behaviors and error recovery mechanisms are built-in. Baxter can perform simple tasks like packing its own gears in boxes.

Rethinking Materials:
Rodney Brooks highlights the use of new materials like powdered metal gears and plastic-impregnated glass for gearboxes in industrial robots. These materials are more affordable and precise than traditional materials.

Computation Replacing Precision:
Brooks emphasizes the role of computation in replacing precision in robotic systems. Using computational models and control theory, it’s possible to use cheaper gears and components while maintaining performance.

The Right Metrics for Robots:
Traditional industrial robots focus on metrics like precision, repeatability, and speed. Baxter’s metrics are different: adaptability, flexibility, ease of use, safety, inexpensiveness, and throughput. Convincing people that these new metrics are important is a challenge.

Built-in Error Recovery and Common Sense:
Baxter’s design incorporates error recovery at all levels, optimizing performance in expected cases and providing robust solutions for unexpected situations. The robot’s behavior-based intelligence coordinates vision and control algorithms, resulting in a basic level of common sense.

Avoiding Edge Case Programming:
Baxter’s common sense eliminates the need for extensive edge case programming. The robot can understand and adapt to changes in the environment, such as missing objects or unexpected obstacles.

Different End Users and Styles:
End users in factories have different styles of using machines, ranging from small business owners who want simple instructions to line workers who want to understand the mechanics. Designing the robot’s interface to accommodate these different styles is important.

Avoiding Outsourcing Hard Engineering Problems:
Brooks cautions against outsourcing complex engineering problems to the end user. Overburdening the user with technical details can make the robot less enjoyable and less efficient to use.

Blind Pick and Place:
The robot, Noelle, demonstrated a blind pick and place task without using vision. Noelle successfully grabbed an object, inferred a pickup location, and released the object at a designated put-down location. The robot adjusted its kinematic posture to avoid collision with another arm. A force field around the robot prevented collisions with other objects.

Training and Execution:
The user, portrayed by Noelle, trained the robot to perform a pick-up and put-down task. The robot learned to infer the pickup location and execute the task successfully. The training process was visually displayed on a screen, showing the robot’s inferred pickup and put-down locations.

User Interface:
The robot’s user interface is designed to understand the user’s intentions and execute the desired tasks. Simple tasks are easy to train, allowing users to quickly teach the robot new skills. More complex tasks can be trained through deeper menus, providing users with advanced customization options. The robot assumes default settings for certain tasks, such as transfer height, but users can adjust these settings if necessary.

Training a Robot to Pick and Place Objects:
The robot is shown in a demonstration where it is tasked with picking and placing objects from a conveyor belt into a box. The robot’s training involves showing it the objects it needs to pick up and where to place them. The robot’s programming allows it to adjust its height and other parameters for different tasks.

Visual Training and Object Recognition:
The robot uses cameras in its hands to visually inspect objects and distinguish them from the background. It autonomously determines the appearance of objects and confidently identifies them after training. The robot is able to recognize objects even with specular reflections.

Customizable Pick and Place Actions:
The robot can be trained to pick up objects from specific areas and place them in designated locations. It is capable of performing both drop and put down actions based on its training. The robot can be instructed to pick up and place different types of objects using various methods.

Fine-tuning Parameters and Advanced Interface:
The robot’s interface allows for fine-tuning of parameters such as speed and height. It includes menus for setting numerical values and adjusting various aspects of the robot’s behavior. The interface also enables deeper control and customization of the robot’s actions.

Safety Features and Limitations:
The robot gently drops its arm when power is disconnected, ensuring safety in case of power failures. The current version of the robot’s software has limitations and cannot perform certain tasks, such as steering a yacht. Future software updates will expand the robot’s capabilities, including machine tending and more advanced tasks.

Adaptability and User-Friendly Task Representation:
Baxter adapts to changes in the world through vision and various sensing options. Users can define tasks without programming sequences, making it easy to modify tasks.

Gripper Installation and Customization:
Baxter comes with a gripper kit, including electric and pneumatic options. Users can easily install different fingers and fingertips on the gripper. Baxter can automatically recognize and compensate for new finger designs through visual feedback.

Elimination of Unnecessary Industrial Robot Features:
Baxter lacks certain traditional industrial robot features, such as fences, key switches, and large emergency stop buttons. This simplifies operation and eliminates the need for extensive safety measures.

ROS-Based SDK and Research Potential:
A ROS-based SDK will be released early in the new year for research and development purposes. The relatively low cost of Baxter opens up research opportunities to a wider range of institutions, including high schools and undergraduate programs.

Impact on Manipulator Research and Applications:
Baxter’s affordability has the potential to transform manipulator research and applications. It enables research beyond manufacturing domains, leading to innovative and exciting uses.

Recognizing the Potential for Both Practical and Playful Applications:
Baxter’s capabilities extend beyond serious research and industrial applications. The speaker acknowledges the possibility of playful and humorous uses, such as the robot performing the Macarena dance.

Understanding User Needs:
Rodney Brooks emphasizes the importance of understanding user requirements when developing the robot. The goal is to create a robot that meets the practical needs of manufacturers, enabling them to use it in their factories and research.

Product Focus:
The robot is not intended as a research tool but rather as a real product, manufactured in large numbers in the USA. Brooks aims to use this robot to transform manufacturing practices, albeit in small ways.

Shankian Influence:
Brooks addresses the suggestion that the robot’s design resembles Shankian scripts, a concept from artificial intelligence research. He clarifies that the robot does not rely on scripts or elaborate knowledge structures. Instead, it utilizes a collection of simple, low-level behaviors that run continuously.

Default Behaviors and Customization:
The robot incorporates default behaviors that prevent collisions between its arms, ensuring safety and smooth operation. While these defaults are designed to be effective for simple tasks, users have the flexibility to modify and rearrange them if desired. This approach is more user-friendly compared to requiring users to understand complex concepts like quaternions.

Customer Feedback:
Brooks highlights the extensive customer engagement efforts, with hundreds of customers visiting factories to evaluate the robot. This feedback is crucial for refining the robot’s design and ensuring that it meets the needs of manufacturers.

Business Challenges in Robotics Development:
VCs’ conflicting advice and demands can lead to disagreements within the company about the robot’s intended purpose. Some VCs insist on complete human replacement by robots, while others prefer partial automation to enhance human productivity.

Integrating Robots into Existing Workflows:
Customers are adapting their production lines to incorporate robots for specific tasks, increasing overall productivity. Collaborative robots perform tasks like moving goods, while humans handle dexterous tasks, dividing the workload efficiently. Simple engineering solutions, such as building fences or alignment tools, enable robots to operate smoothly.

User Interface and Feedback:
Engineers initially believed users would naturally show the robot how to move objects and identify them on the screen. However, users found this process unnecessary and redundant, leading to a simplified interface where objects are learned without manual outlining. User feedback has highlighted the need for intuitive and user-friendly interfaces.

Teaching Force Application:
The current version of the software lacks force application capabilities, but methods are being developed to teach robots how much force to apply. The current hand allows users to set a threshold for grasping force, though this feature will be added in a later software version. Challenges exist in teaching force application, especially for delicate tasks involving glass or buttons.

Consumer Perception and Managing Expectations:
The perception of a robot’s capabilities depends on its advertised purpose. A single-purpose robot like the Roomba will have lower expectations compared to a customizable development platform. Setting realistic expectations for consumers is crucial to avoid disappointment. Rethink Robotics conducts seminars to inform potential customers about the robot’s limitations.

Foxconn’s Robot Ambitions:
Foxconn’s claim to replace workers with 3 million robots is viewed as unrealistic. No evidence of Foxconn’s robots exists, and companies approached by Foxconn cannot meet their demand. Foxconn’s attempt to reduce salary demands is seen as a driving factor behind this claim.

Longevity and Future Prospects:
Rodney Brooks acknowledges that the current robot may be too early in its development and might not succeed. However, he believes that robots like this will be ubiquitous in the future, though other companies may capitalize on their lessons learned.

Cost-Effectiveness:
The robot’s operating costs are estimated to be around $3 per hour, including electricity and other expenses. The payback period for the robot’s investment depends on its specific application and usage.

Baxter’s Speed and Role:
Baxter is slower than humans but can be cost-effective if the task is repetitive and affordable. It is not meant to replace humans but to assist them by automating tasks, freeing up time for other activities.

User Training and Feedback:
Baxter’s user interface (UI) is being designed to help users identify cases where the robot cannot perform a task. Currently, a national distributor network provides support and feedback, but the goal is to eventually sell Baxter online without intermediaries. Data collection from Baxter is not currently enabled due to sensitivity among potential customers.

Software Updates:
Software updates for Baxter are done via a thumb drive, rather than direct network access.

Baxter’s Limitations:
Baxter is a stationary robot and cannot move. Its mobility was restricted due to legal obligations related to Rodney Brooks’ previous role at iRobot.

Training and Task Transfer:
Once a task is trained on one Baxter robot, it can be transferred to other Baxter robots using a thumb drive. Network-based transfer will be enabled in the future.