By Prof. Wayne Book and Hannes Daepp, Georgia Institute of Technology
–From the CCEFP Newsletter
1.) a means for demonstration of the applicability of research projects to real world applications and
2.) a guide the formation of additional research projects to meet the challenges seen in implementing relevant portions of the test bed’s functionality.
In addition, they illustrate the potential of fluid power to prospective students and excite their imagination for a career in the related areas. The CCEFP test beds are selected to collectively cover a broad range of power levels and Test Bed 4, the Compact Rescue Robot, represents applications in the 100W to 1 kW range, roughly human scale applications. In this range you will not find many current fluid power applications on the market and contrary to applications at a higher range of power, such as the excavator and passenger vehicle, the Center elected a more exotic application: a walking rescue robot with a large number of degrees of freedom and a limited market at this time. While this has endured some scrutiny from those involved in more conventional applications, a rescue robot epitomizes many challenges that can be found in this power range, and illustrates opportunities for some new products in the fluid power industry. Applications are envisioned in related areas such as service robots, assistive devices and construction and agricultural applications. The closest relevant devices employing fluid power at or near commercial availability include the Big Dog Robot1 by Boston Dynamics for rough terrain transport and the Bear Robot2 (battlefield extraction assistance robot) by VECNA Robotics that would indeed profit from improved compactness and efficiency.
The challenges foreseen for the CRR (Compact Rescue Robot) include efficient small scale generation of power, either pneumatic or hydraulic, effective control algorithms, especially for pneumatic servo control, and effective operator interfaces that must be substantially different than those for larger applications where the operator tends to be riding on the device.
The means of mobility a major decision point of CRR design. Why legs? In a rescue situation, it is anticipated that unstable debris, damaged stairways and obstacles in the path will be encountered. This is the situation reported in the Fukushima Dai-ichi nuclear reactor where four iRobot military robots of two designs have been modified for exploration of the high radiation areas of the plant. While rescue of victims is not the mission of these robots, gaining access to the points of interest in the plant is potentially very similar, since a hydrogen gas explosion in the plant has resulted in significant damage to the buildings. The PackBot and larger Warrior robots are treaded vehicles and operators report difficulty in climbing stairs, gaining traction, opening doors, and keeping upright.3 Operator experiences bluntly placed in a blog by one of the operators yields a realistic look at the challenges of operating a robot in a disaster scenario. A legged, compact, pneumatic or hydraulic robot would be able to address some of the issues that challenge the operators there.
Electric rescue robots are most commonly tracked or wheeled vehicles. Negotiating stairs and rugged terrain presents challenges for these “continuous contact” designs that legged locomotion does not typically encounter. Legs can move from one stable contact point to another without contact with unstable regions in between. Furthermore, intermittent, reciprocating actuation of legs by pneumatic or hydraulic cylinders is more common than would be for electrical drives which are inherently rotational.
With these and other issues in mind, Keith Wait and Michael Goldfarb of Vanderbilt University designed and constructed the current CRR. The CRR is a four legged, pneumatically actuated vehicle. Each leg consists of three joints with independent position control of each joint. The flow of air to the joint is modulated by custom, digitally controlled valves rated at 300 psi and 300 deg. F. Each joint has its own PIC based controller that receives commands from the operator via CANbus and feedback from joint and valve positions. At Vanderbilt, our CRR was operated with several different standard gaits, such as walking and trotting. These gaits were choreographed in advance and can be used for locomotion over more or less even terrain, which is not seen as the more valuable means of locomotion in a rescue scenario. We believe that critical operations in a rescue scenario will demand precision placement of the legs and the legs may also be used for remote manipulation commanded by the operator. A remote operator is deprived of many of the normal cues of a machine mounted operator. Some of these cues might be provided to the operator virtually in direct substitution of normal cues while the need for others can be replaced automatically by an intelligent vehicle control.
Since many electrically driven, battery powered devices currently exist, natural bench marks were available. We were able to use a database compiled by the National Institutes of Standards and Technology (NIST)4 which gave relevant information on dozens of experimental and commercial vehicles that chose to participate in its evaluation program. The CRR is targeted to meet or exceed the capabilities found in the NIST guidebook for response robots, and to provide other capabilities, such as high lifting force, as well.
In addition to the vehicle itself, an operator’s workstation is essential to the missions envisioned for a human scaled robot. The prototype has been constructed at Georgia Tech. It consists of two haptic manipulators that will enable the operator to feel what the robot feels on its front legs and direct both manipulation and locomotion activities with natural sensory feedback. Vision can be provided through conventional head mounted displays as well as an experimental “trailing view” of the robot drawn onto a picture taken a few moments earlier. The work on this display has been undertaken in conjunction with Kyoto University and is separately funded5.
Providing pneumatic power to a small, untethered robot could be tackled with a conventional IC or electric engine driving a pump. Small IC engines are extremely inefficient and must either idle or stop and start. Project 2B.16 at Vanderbilt University, lead by Eric Barth, has produced a free piston engine that directly drives a compressor that already shows advantages over the conventional approach. It can instantly restart as flow is needed. It is quiet running and needs no transmission connecting it to a compressor. Compared to batteries it provides longer operation for the same weight, since the additional mass needed to extend the time is only the mass of the fuel, not the mass of additional batteries. The current version is being revised to increase pressure and flow delivered and will be mounted on the CRR for demonstration in the first quarter of 2012.
Projects 3A.17 and 3A.38 focus on the operator interfaces of fluid powered equipment and specifically on two test beds: the excavator and the CRR. A major disadvantage of legs, particularly a four legged vehicle, is that the balance of the robot depends on placement of the legs moved previously. Project 3A.29 addressed the challenges of the pneumatically powered rescue robot joints. The interaction via passified pneumatic control showed enhancements over conventional PID type controllers.