TRIUMVIRATE. dormouse. Final Report. David Kim, Erick Lang, BinhAn Ta 5/2/2008

Transcription

1 TRIUMVIRATE dormouse Final Report David Kim, Erick Lang, BinhAn Ta 5/2/2008 This document is the final report on the project dormouse: a project to develop an accelerometer based solution which addresses problems found in traditional computer mice currently in the market. By using an accelerometer, wireless transceivers, microcontrollers, and few buttons, we were able to create a mouse that resolves all the problems identified with the traditional mouse. The design choices, challenges, and solutions to the challenges for building such product are explained in this document. Also, production costs and possible improvements are explored in depth. 1 P a g e

2 Contents List of Figures... 3 List of Tables... 3 Introduction... 4 Design Concept... 4 Analysis of User Related Design Choices... 4 Comfortable feel... 5 Ease of use... 5 Portability... 5 Analysis of System Design Choices... 5 Position Module... 6 Action Module... 7 System Module... 7 Alternative designs... 8 Addition of Gyroscope... 8 Future Works (Afterthought)... 9 Conclusion... 9 Appendices Cost analysis Challenges Integration PID Controller References P a g e

3 List of Figures Figure 1. Modules... 6 Figure 2. Verlet Integration Figure 3. Verlet Integration with Dampening Figure 4. PID Controller List of Tables Table 1. Cost Analysis P a g e

4 Introduction In today s computing world, mouse is an indispensable tool. With their functionalities to map the user movements into two-dimensional space and quickly select and/or open menu items have brought great convenience to the end users. We as users have become much familiarized with the concept of mouse and cannot imagine working with today s PCs without them. However, there is still some room for improvement that needs to be addressed. First issue that needs to be addressed is the requirement for extra desk space. Second is that the traditional mouse requires users to move their hands away from the keyboard in order to operate the mouse. Although in most scenarios this extra movement is not critical factor in performing any tasks, it is nevertheless an excess movement and should thus be eliminated if possible. Third is the requirement for wires, also eating up desk space and limiting portability. This issue is already addressed in the today s world in some situations with wireless Bluetooth mice, but still remains an issue with most traditional mice. dormouse tries to address these three issues and improve upon them. dormouse seeks to deliver a futuristic mouse which is wireless, on-keyboard, and compact. After identifying the problem we wanted to solve, we spent some time discussing about the core design of the system. We identified that most mouse users do not use one of their thumbs -- they use one thumb to press the spacebar, but the other one is often unused. If we could utilize the unused thumb to control mouse, then the users will no longer need to move their hands away from the keyboard. Also, to allow the users to be able to click button while having the mouse pointer stable, we decided to develop a buttons device on the opposite hand of the position thumb. We decided to call each device, position module and action module, respectively. The separation of the functions is what made dormouse feasible. Without each function falling onto different hands so nicely, dormouse would never have seen the light. With these core ideas in mind, we have designed and developed dormouse. Design Concept This section provides the analysis of design choices we have made and explains the basic concepts of how dormouse operates. Analysis of User Related Design Choices In designing dormouse, we considered the followings to be important to the users: Comfortable feel Ease of use 4 P a g e

5 Portability Comfortable feel As in any other mouse, there must be a way to click to select, drag and drop, and open menu options. For this purpose, our design includes the use of five tactile buttons which provide feedback when clicked. Also, this type of button is familiar to everybody that uses computer nowadays since this is what virtually all computer mice use as input. Ease of use To adapt to the limited movement of the user s thumb, and to the user s choice of cursor speed, we decided to provide a way to set sensitivity of the position. Some users would want to move the cursor fast, whereas some would move slower. The sensitivity buttons exist to provide extra degree of user freedom and ease of use. Another important design decision was that of having the buttons go on the opposite hand as the accelerometer. If the buttons were designed to be on the same hand as the accelerometer, the cursor would move whenever the user tried to click on an object on the display, which would result in great user frustration. Portability For portability, dormouse was designed as a wireless solution. This way dormouse can be operated from different distances from the computer, going as far as 20 meters. The wireless solution also allows users to carry around the device with ease. Analysis of System Design Choices After looking at the previous requirements, we chose the system design that best fitted such objectives, which led us to divide dormouse into three main components: Position module Action module System module 5 P a g e

6 User will wear the modules as in the following diagram: Figure 1. Modules Position Module This module detects the acceleration changes from the user's thumb with a tri tri-axes accelerometer. The accelerometer's signal is changed to bits through an analog to digital converter module that is included in the microcontroller. Then the signal is sent to the computer through the use of a wireless protocol SimpliciTI, created by Texas Instruments. SimpliciTI is a light-weight weight protocol that does not require extra hardware. There are six API calls for this protocol: SMPL_Init - prepares the hardware for wireless communication SMPL_LinkListen - waits for other devices to connect SMPL_Link - connects to a network SMPL_Send - sends message to wireless ne network SMPL_Receive - receives message from wireless network 6 Page

7 Although the microcontroller used can only send wireless signals at a slow speed, such signals can still be sent is the same frequency as any other computer mice currently in the market. Most PS/2 mice operate at a frequency of 200Hz; this is still achievable with SimpliciTI protocol. Action Module The action module makes use of the five tactile buttons with the following functions: left click, right click, sensitivity up, sensitivity down, and reset. When implementing buttons, there are two possible ways of reading the input signals: polling, or interrupt handling. Using polling wastes more power, because it constantly reads the input line. Interrupt based input, on the other hand, saves on power consumption by making the microcontroller read the input only when the signal has changed. We decided to implement an interrupt based input for the buttons in order to obtain a low power consumption. To do so, we had to implement the action module with two microcontrollers, because the MSP430F2274 only has the interrupt port 2 available for development, and these pins were already taken by SimpliciTI protocol. Two MCU solution solved the problem by having one MCU for reading the button interrupts and sending UART signals to the second microcontroller, which wakes up on UART receive and sends the signal wirelessly. As explained in the user related design choices section, the sensitivity buttons were designed into our prototype for the user's comfort and in regard to the physical limitations of the thumb's movement. As in any other traditional mouse, the left and right buttons are used for selection, and menu display, respectively. The reset button is present to provide a mean to readjust the mouse pointer to the center of the screen. Without the reset button, any movement of the thumb (or hand) would be registered and eventually the user could lose control if the user were to move. System Module This module receives the signals sent from the action and the position module by wireless communication, and transfers the data to the computer through serial port (USB). After this, an application reads the USB port and does the following operations on the data: 1. Takes the integral of the data to obtain velocity and position from the acceleration signal 2. Runs algorithms on the data in order to reduce/cancel unwanted acceleration changes that may be present from tilting or noise 3. Maps out the result on the screen 7 P a g e

8 Alternative designs This section discusses the alternate designs we considered but rejected. They are presented along with reason for their rejection. Casing One of the design considerations that came under most scrutiny was the case design. Due to having no mechanical or structural engineer who has had experience with designing casing for a mobile device, we spent some time researching some materials. First we decided that the casing should be done with plastic like most mobile devices. We considered using CAD (Computer Aided Design) tools to design the casing in three-dimensional space and send it into a workshop to get it fabricated. During the second semester we decided to reject this idea because no member of the team had any experience with CAD tools let alone the access to them. Learning will be too time consuming and CAD just did not fit well with development of a prototype device. This was why we adopted the idea of Sculpey. We wanted something that was easy to shape, yet provide the structural strength to be a casing for our demonstrations. Also, not having much prior experience in such materials, we wanted to make sure we chose a medium which is easy to shape but still used popularly in professional settings. As it turns out, Sculpey was not a bad decision, but still required too much effort and time to someone who did not have any prior experience working with them. The thinly shelled case changed in its shape during the baking process and the final product fell apart upon slightest abuse. In the end, we turned to glove-based design for casing. This worked very well, giving us confirmation that glove-based designs was the decision we should have made at the start of the project. The glove-based approach is further discussed in the Future Works section. Addition of Gyroscope At the beginning of the project, we discussed about adding a gyroscope (or other MEMS devices) to stabilize the signals or the device itself. However, this idea was rejected upon further discussions for the following reasons: Limited number of inputs on Texas Instrument s MSP430. Unwillingness to add extra hardware cost. Our firm belief that accelerometers can deliver every need for our system. 8 P a g e

9 Afterthought, we are still undecided on our decision in this. We were able to obtain very clean position reading from using only the accelerometer -- good enough to navigate to a point of interest and hover over it for an extended period of time. However, from what we learned from an online article (Melanson) two weeks prior to project deadline, Microsoft had decided to develop a similar device with accelerometer, gyroscope, and magnetometer. If Microsoft wants to use extra devices, perhaps addition of gyroscope would have aided stability of our readings. Future Works (Afterthought) dormouse s design can be improved to increase the accuracy and enhance the efficiency. First, PID controller mechanism which is used to eliminate tilt acceleration can be replaced by using a gyroscope. A gyroscope is used to measure angular velocity. Thus, by integrating the angular velocity, the change of angle can be found with accuracy of ± 0.02 degree. As a result, the tilt angle can be measured and eliminated with higher accuracy and faster processing speed (determining tilt angle with gyroscope involves less computation compared to PID controller mechanism). Second, a pair of gloves can be used to hold the device instead of using cases made from Sculpey. Gloves are flexible to put on, cheap to obtain, and easy for users to use. In fact, we have had best results with the dormouse s final prototype mounted on a glove. Third, a power switch can be used to turn off the microcontrollers and accelerometer when these devices are not in use. This will improve the life time of our device. Fourth, a device driver can be implemented to replace our Java program. This will lead to better efficiency since it does not require any attentions from the users to make the dormouse operate properly yet provide faster and more reliable control. Lastly, Bluetooth communication protocol can be used to replace the SimpliciTI protocol since Bluetooth protocol is widely used and is more efficient, not to mention the extra security provided from its frequency-hopping protocol. Conclusion In the beginning of this report, we showed that an accelerometer based mouse has benefits over the traditional computer mice. We have implemented such mouse and demonstrated the feasibility of these devices with simple implementation and low cost. Although, there are many improvements that can be made to dormouse as discussed in the future work section, we believe dormouse, with its innovative design, has the potential to be a great product in the current market. 9 P a g e

10 Appendices This section discusses miscellaneous aspects of the dormouse project. Cost analysis Parts List Quantity Unit Cost Total Cost Prototype Mass Production Prototype Mass Production EZ430 Target Board 2 $10 $5.55 $20 $11.1 EZ430 Programmer 1 $20 $8.67 $20 $8.67 MSP430 MCU 1 $10 $1.9 $10 $1.9 ADXL330 Accelerometer 1 $35 $5.45 $35 $5.45 Lithium Coin Battery 2 $1.90 $0.54 $3.8 $1.08 Battery Holder 2 $1.95 $0.39 $3.9 $0.78 Buttons 5 $.50 $0.12 $2.5 $0.6 Miscellaneous (wires, pin N/A $7.8 $3.76 $7.8 $3.76 headers ) TOTAL $103 $33.34 Table 1. Cost Analysis Currently, the average price for a Bluetooth wireless mouse is about $60. If a dormouse device were to sell at the price of $50, it would still be $10 dollars cheaper than a Bluetooth wireless mouse, with a profit of $ The cost for materials used in mass production can be reduced further because all the prices include in the part list of mass production based on price for bulk of Challenges This section describes most important challenges that we overcome to make dormouse feasible. Integration Throughout the project, our main concern has been the accuracy and stability of the calculated position. Due to the nature of double integration, any error in acceleration accumulates exponentially. This exponential error has been what made Cornell student s Minority Report Mouse project (Kohli) fail. 10 P a g e

11 At first, we decided to use Euler s integration a method implemented as simple addition of samples each time slice. This method failed as it did not account for any in-between values and those values were necessary for clean integration results. However, our research into the subject matter found a brief mention of Verlet integration from the Cornell s project. Verlet integration is an advanced integration method which calculates position straight from acceleration by utilizing the previous sample values and the values previous to the previous. The original Verlet integration, which predicts next position of a molecule from the given acceleration is given as, Figure 2. Verlet Integration By subtracting delta t from all terms, we can find the current position from the past accelerations. In our project, we implemented a modified version of Verlet integration that incorporates dampening of velocity into the equation. It is shown below: Figure 3. Verlet Integration with Dampening It was implemented with the following Java code snippet. // Verlet integration with Velocity Dampening //x(t + \Delta t) = (2-f) x(t) -(1-f) x(t - \Delta t) + a(t)(\delta t)^2 pos_x = (2.0 - damper) * prev_pos_x - (1.0 - damper) * prev_prev_pos_x + prev_accel_x * dt * dt; // Now set the current to previous so current can be used as previous next iteration prev_prev_pos_x = prev_pos_x; prev_pos_x = pos_x; prev_accel_x = accel_x; Verlet Integration, along with the PID controller, made the position module possible. PID Controller Another technical challenge we had was filtering out the tilt from the movement acceleration of the position module. We realized sometime into the development that it is impossible to detect tilt and movement simultaneously with just an accelerometer. This is usually not a 11 P a g e

12 problem, but considering the movement range we detected was within +-2g, even a slight tilt such as 15 degrees tilt resulted in 2% error. This 2% error is important in that it is asserted constantly as long as the accelerometer remains tilted relative to the starting tilt, and the constant assertion results in exponential error accumulation. In order to rid of this tilt, the position module must slowly readjust itself to the gravity reading. The slow adjustment was implemented with a PID controller. PID controller s operation is summarized in the following diagram. Figure 4. PID Controller PID controller uses proportional, integration, and derivative errors to adjust a measured value to a set value. In dormouse s case, we used acceleration read to be goal value and adjusted gravitational setting toward the acceleration to make gravitational reading match the acceleration read over the course of movement. In our design, usual movement was considered to last 0.5 second, and P value was chosen to be 63/64. This value was selected for its property of, (63/ 64) 100 Which is a reasonable adjustment speed that does not kill the movement signal but gets rid of tilt P a g e

13 References [1] H, Weinberg. "Analog Dialog." Ask The Applications Engineer-29. Sep Apr 2008 < >. [2] Liu, H.H.S.; Pang, G.K.H., "Accelerometer for mobile robot positioning," Industry Applications, IEEE Transactions on, vol.37, no.3, pp , May/Jun 2001 [3] "CTM: PID Tutorial." Michigan Engineering. 26 Aug University of Michigan. 2 May 2008 < [4] "Verlet Integration." Wikipedia. 07 Apr Wikipedia. 2 May 2008 < [5] Kohli, Aseem. "Accelerometer Mouse." Computing at Cornell Home Page Cornell University. 2 May 2008 < M249_AK288/index.htm>. [6] Melanson, Donald. "Microsoft patent application shows off wearable mouse concept." Microsoft. 17 Apr Apr 2008 < 13 P a g e