Proceedings of the Multi-Disciplinary Senior Design Conference Page 7

Project Number: P13202

TigerBot Humanoid Hip SErvo

Matthew LeStrange
Electrical Engineering / Vu Nguyen
Electrical Engineering
Brandon Baker
Mechanical Engineering / PJ Haasenritter
Computer Engineering

Proceedings of the Multi-Disciplinary Senior Design Conference Page 7

Abstract

A major weakness in the past RIT humanoid robot projects is in the high torque hip and knee joints. The goal of the team is to design a high torque servo for humanoid robots with a focus on the hip joint, while maintaining a low cost compared to similar robotic servos on the market. This prototype is designed to provide variable speed and high torque output configurations and has several modular features that allow the servo to be designed into almost any robotic joint. The design also includes a smart communication interface to send feedback and receive commands between the servo and the robot controller.

introduction

The end goal of the project is to create a reconfigurable servo capable of serving in the high torque joints of the TigerBots (a set RIT senior design project, which are a humanoid robot approximately three feet tall). The servo has a universal housing and mounting system to allow the servo to be used in almost any orientation. More specifically, the primary design criteria came from the goal of using the servo in the hip joint. The hip joint was a point of failure in the first two TigerBot designs. The TigerBot projects have been a part of the multidisciplinary senior design program where the team has been given the goal was to design a two to three foot tall humanoid robot, during the twenty weeks of the courses. The TigerBot ideally would be able to walk, pick it’s self up after a fall and perform other complex movements. The primary area of weakness in these projects was the hip joint. The hip joint typically consists of three servos to provide three degrees of freedom.

When the robot attempts to walk, one hip must be able to lift the entire leg and the other hip must be able to support the entire weight of the robot. The very first area of investigation for the design of the servo was to look at the problems in the previous robots. The main problem with the first TigerBot design was that the robot had a heavy torso with all the electronics and all the weight was placed on the weak servos in the hips. This caused a large side load on the servo shaft. The robot was unable to stand, until an additional bracket was added. This robot also has legs that were longer than human proportions and the knees were similarly very weak. This robot used two types of Hitec servos in the hips that provide up to 416 oz-in of torque at 7.4V.

The second TigerBot design is based more on closely on human proportions. This design is able to stand on its own, make some small movements and balance itself against external forces. Walking for the second TigerBot is still very difficult because the batteries were kept in the feet and the weight of the robot torso was still very high. This robot used RoBoard servos that produced up to 486 oz-in at 7.4V. There was a discussion during the second TigerBot’s design process about designing a servo for the robot, but it was decided that would have to become its own future design project and that is part of where this project came about.

After examining the design of the previous two TigerBots and having a strong understanding of the issues with the servos in those designs, several servos on the market were examined to understand the types of servos available on the market and what aspects and features should be included in the design of the hip servo. The first TigerBot team used Hitec servos, which are a brand typically used in hobby application and did not provide enough torque since the need was to design to have a minimum torque of 500 oz-in. One major positive about the Hitec servos is that they have many mounting features, so that they can be used in a variety of orientations for more different joints without additional weight from brackets.

The second TigerBot team used RoBoard RS-1270 servos. These were very small (40.4 x 20.1 x 40.3 (mm)) for the amount of torque they can produce (486 oz-in). However, these servos were difficult to mount and extra brackets were needed to hold them in place.

There is a servo manufacture, Dynamixel, that makes many different models of high torque servos, but they are very expensive ($300 to $500). Dynamixel is one of a few other companies make what they call “smart servos” meaning that they use digital control and are capable of sending feedback to the servo controller, microcontroller or processor that is sending the position commands. The Dynamixel servos use serial TTL logic or RS485 multi drop to send position and other variable commands. Dynamixel servos have information on the position, velocity, torque, current, voltage and temperature. The PID control loop gains can be read or adjusted through the serial interface. Dynamixel servos have very high torque (up to 1000 oz-in or more) and rows of mounting holes, but the downside is the cost. For a humanoid robot, there would typically be twenty or more servos and the cost for even ten Dynamixel servos would exceed the desired budget of the robot. Another difference with the Dynamixel servos is that they are typically designed to operate at in the 11.1V to 14.8V range.

The last servo that was studied in-depth in the initial research process was the XQ Power Servo (XQ-S5650D). This servo was found by the current TigerBot team and is being used in their design. These servos do not have the smart features, but can produce high torque (666-833 oz-in) and are relatively cheap ($100 to $150). The major problem with theses servos is that they only have 4 mounting holes, so addition brackets and material have to be added to incorporate the servo into the robot link. These are also a relatively new product and several boards failed in these units during the robot testing.

The goal is to take the best features from the servos studied and add some unique additional features. The design will focus on having a high torque output (approximately 700 to 800 oz-in) and include mounting holes on all sides, so no additional brackets are needed. The gearbox design will also be modular in that it will have three gear ratios designs all using the same enclosure to allow three different torque/speed combinations. The design will also feature a unique optional dual output shaft (see Fig. 3) that provides a method to create stronger joints. The interface will feature a standard PWM servo control interface and additionally a CAN bus interface. The CAN bus allows up to 31 servos to be connected on a single bus to the robot controller. This smart interface provides their position commands and allows additional feedback to be sent between each servo and the robot controller. Cost is another critical design requirement, staying within a $750 budget and designing a servo that could be potentially be produced (at a quantity of 100) at a cost of less than or those on the market. Based on this the goal is to be able to manufacture the servo for approximately $100.

process

The initial design needs and specifications were left very open for a solution to the failure in the robotic hip joints. It was suggested that a new servo be designed, but essential the goal was to design a solution for the high torque joints that had failed in previous designs. There was an idea of designing a system that would act as all three axes of the hip joint, but it was decided this was too complex and would not easily be adapted to other joints.

The customer needs for the servo are very simple; the main concern is the need for the servo to produce enough torque for the hips and other high torque joints and remain at approximately the same size as other servos. The next sets of needs are the design must modular in the gearbox for different speed/torque capabilities; the case should also be modular in that it can be used for different joints. Some robot links would perform better if the output shaft went through the entire servo. For example, in a robotic knee joint there is one degree of freedom and the dual output shaft would be able to attach to two points in the joint and reduce the effect of external forces. For the electronics, the servo needs to use PWM to control the position of the servo. PWM is the standard control method of servos; typically pulses from 1ms to 2ms at 50 Hz define the angular position. There is also a need to have feedback to the robot controller to give the robot the position of the servo and the amount of current being consumed, which can be related to the torque.

The specifications of the servo were based on benchmarking against some of the other servos discussed in the Introduction in addition to the needs of the customer. The size was designed to be comparable to those servos with the largest dimension being less than 2.5 inches. The specification for the torque was one of the more complicated numbers to come up with. By making the servo larger, it is easy to increase the torque because a larger motor and higher gear ratio can be accommodated. The final specifications for toque were set based on the size/weight. The specification is 100 to 200 oz-in per oz and 125-200 oz-in per in3. The speed of the servo was calculated from video of people walking and determined the motion of the leg needs to be 25-40ᵒ/second. To accommodate the modular speed/torque ratio need, there was a specification to design three gear boxes for the servo.

The electrical supply voltage specification came from the standard 7.2V and 7.4V used for the batteries in past years. The goal was to function from the 6V to 8V range as a minimum. The system also needed to provide position and current feedback and those required specifications based on benchmarking against the Dynamixel smart servos. The position should be accurate to at least 1ᵒ and have a precision of 0.25ᵒ. The current should be measured a range from 0A to 5A at an accuracy of 5 to 50mA. Another feature of the smart control is the servo can make the motor brake in addition to clockwise and counterclockwise motion.

A block diagram of the system is shown in Figure 1 to give a general overview of the different components of the system and how they interface. This diagram is based on the final design, but it contains each of the major components whose concept and design will be discussed in the rest of this section.

Figure 1: System Block Diagram

The mechanical housing must provide space for and mounting space for all the components inside the servo. The enclosure design must also provide several mounting interfaces, so the robot can use the servo joint can operate in any direction. Looking at Figure 2, the CAD model of the servo has 4 mounting holes on four sides of the servo, so there are a variety of orientations that can be used. The lack of mounting features was an issue with many of the servos studied in the initial design phase.

Figure 2: CAD Model of Servo

The majority of servo motors enclosures are made of a plastic or a larger portion of plastic and material consideration was important in this design. Because this is a design and prototyping project, the material had to easy to machine and relatively cheap. One option would be to rapid prototype a plastic enclosure using 3D printing, but this can be expensive and it is difficult to rework and change the design. Aluminum is a cheap and easy to machine material and it also provides thermal dissipation for the electronics and motor.

The mechanical enclosure was machined both by hand and on a CNC machine. Initial testing was done using foam, particle board and Lexan (a clear polycarbonate). These materials were faster to machine, in some cases minutes instead of hours, so the design and CNC G-code could be verified before producing the final product.

The gearbox design is one of the more complicated portions of the servo. Not only do the gears all have to fit in a small space, but the gearbox should also be modular. The typical servo gear layout used in every servo we studied was the same (see Fig. 3), so our design copied this gear train layout. A cutaway view of our servo is shown in Figure 4 and shows the gears train used in our design.

Proceedings of the Multi-Disciplinary Senior Design Conference Page 7

Figure 3: RoBoard RS-1270 Servo Gearbox

Figure 4: CAD Model Cutaway View

Proceedings of the Multi-Disciplinary Senior Design Conference Page 7

Figure 5: TigerBot Servo Compound Gears

The small gears used on the motor were simply cut to size from stock and put in a lathe to drill the center hole the motor output was pressed into. Similarly, the last gear, attached to the output shaft, was pressed into place after precisely drilling out the hole and cutting the shaft down to a suitable press-fit diameter.

The three different gear ratios are achieved both by swapping out gears and changing the motor position. For the 211:1 ratio, which we will call the base configuration, is achieved with one 12 tooth gear on the motor, two 48-to-12 tooth compound gears, on 48-to-15 tooth compound gear, and one 50 tooth gear. The second ratio, 264:1, is achieved by replacing the motor gear with a 10 tooth gear, and the first compound gear with a 50-to-12 compound gear. The third ratio, 330:1, is achieved by modifying the second ratio configuration by replacing the 10 tooth motor gear with an 8 tooth one and moving the gear closer to the first compound gear. There are two sets of holes for mounting the motor, perpendicular to each other. One set of holes is offset so that the 8 tooth motor gear will continue to mesh with the rest.