Sensor readings from a wall-following robot

### Context

- The data were collected as the SCITOS G5 navigated through the room following the wall in a clockwise direction, for 4
rounds. To navigate, the robot uses 24 ultrasound sensors arranged circularly around its "waist". The numbering of the ultrasound sensors starts at the front of the robot and increases in clockwise direction.

- The provided files comprise three diferent data sets.

- The first one contains the raw values of the measurements of all 24 ultrasound sensors and the corresponding class label (Moving forward, turning left, etc). Sensor readings are sampled at a rate of 9 samples per second.

- The second one contains four sensor readings named 'simplified distances' and the corresponding class label l (Moving forward, turning left, etc). These simplified distances are referred to as the 'front distance', 'left distance', 'right distance' and 'back distance'. They consist, respectively, of the minimum sensor readings among those within 60 degree arcs located at the front, left, right and back parts of the robot.

- The third one contains only the front and left simplified distances and the corresponding class labell (Moving forward, turning left, etc).

- It is worth mentioning that the 24 ultrasound readings and the simplified distances were collected at the same time step, so each file has the same number of rows (one for each sampling time step).

- The wall-following task and data gathering were designed to test the hypothesis that this apparently simple navigation task is indeed a non-linearly separable classification task. Thus, linear classifiers, such as the Perceptron network, are not able to learn the task and command the robot around the room without collisions. Nonlinear neural classifiers, such as the MLP network, are able to learn the task and command the robot successfully without collisions.

- If some kind of short-term memory mechanism is provided to the neural classifiers, their performances are improved in general. For example, if past inputs are provided together with current sensor readings, even the Perceptron becomes able to learn the task and command the robot successfully. If a recurrent neural network, such as the Elman network, is used to learn the task, the resulting dynamical classifier is able to learn the task using less hidden neurons than the MLP network.

- Files with different number of sensor readings were built in order to evaluate the performance of the classifiers with respect to the number of inputs.

### Content

File sensor_readings_24.csv:

- US1: ultrasound sensor at the front of the robot (reference angle: 180°) - (numeric: real)
- US2: ultrasound reading (reference angle: -165°) - (numeric: real)
- US3: ultrasound reading (reference angle: -150°) - (numeric: real)
- US4: ultrasound reading (reference angle: -135°) - (numeric: real)
- US5: ultrasound reading (reference angle: -120°) - (numeric: real)
- US6: ultrasound reading (reference angle: -105°) - (numeric: real)
- US7: ultrasound reading (reference angle: -90°) - (numeric: real)
- US8: ultrasound reading (reference angle: -75°) - (numeric: real)
- US9: ultrasound reading (reference angle: -60°) - (numeric: real)
- US10: ultrasound reading (reference angle: -45°) - (numeric: real)
- US11: ultrasound reading (reference angle: -30°) - (numeric: real)
- US12: ultrasound reading (reference angle: -15°) - (numeric: real)
- US13: reading of ultrasound sensor situated at the back of the robot (reference angle: 0°) - (numeric: real)
- US14: ultrasound reading (reference angle: 15°) - (numeric: real)
- US15: ultrasound reading (reference angle: 30°) - (numeric: real)
- US16: ultrasound reading (reference angle: 45°) - (numeric: real)
- US17: ultrasound reading (reference angle: 60°) - (numeric: real)
- US18: ultrasound reading (reference angle: 75°) - (numeric: real)
- US19: ultrasound reading (reference angle: 90°) - (numeric: real)
- US20: ultrasound reading (reference angle: 105°) - (numeric: real)
- US21: ultrasound reading (reference angle: 120°) - (numeric: real)
- US22: ultrasound reading (reference angle: 135°) - (numeric: real)
- US23: ultrasound reading (reference angle: 150°) - (numeric: real)
- US24: ultrasound reading (reference angle: 165°) - (numeric: real)
- Classes: Move-Forward, Slight-Right-Turn, Sharp-Right-Turn, Slight-Left-Turn

File: sensor_readings_4.csv:

- SD_front: minimum sensor reading within a 60 degree arc located at the front of the robot - (numeric: real)
- SD_left: minimum sensor reading within a 60 degree arc located at the left of the robot - (numeric: real)
- SD_right: minimum sensor reading within a 60 degree arc located at the right of the robot - (numeric: real)
- SD_back: minimum sensor reading within a 60 degree arc located at the back of the robot - (numeric: real)
- Classes: Move-Forward, Slight-Right-Turn, Sharp-Right-Turn, Slight-Left-Turn

File: sensor_readings_2.csv:

- SD_front: minimum sensor reading within a 60 degree arc located at the front of the robot - (numeric: real)
- SD_left: minimum sensor reading within a 60 degree arc located at the left of the robot - (numeric: real)
- Classes: Move-Forward, Slight-Right-Turn, Sharp-Right-Turn, Slight-Left-Turn

### Acknowledgements

These datasets were downlaoded from the UCI Machine Learning Repository

Lichman, M. (2013). UCI Machine Learning Repository [http://archive.ics.uci.edu/ml]. Irvine, CA: University of California, School of Information and Computer Science.

### Inspiration

Use these ultrasound readings to predict the class, i.e. given these readings, is the robot moving straight? turning left?

### 背景

- 本数据集的收集源于SCITOS G5机器人沿顺时针方向跟随墙壁在室内进行四次绕行。为了实现导航，机器人采用环绕其腰部呈圆形排列的24个超声波传感器。超声波传感器的编号从机器人的前方开始，沿顺时针方向递增。

- 提供的文件包含三个不同的数据集。

- 第一个数据集包含了所有24个超声波传感器的原始测量值以及相应的类别标签（例如：前进、左转等）。传感器读数以每秒9个样本的速率进行采样。
- 第二个数据集包含了四个名为'simplified distances'的传感器读数和相应的类别标签l（例如：前进、左转等）。这些简化的距离分别被称为'前方距离'、'左侧距离'、'右侧距离'和'后方距离'。它们分别对应于机器人前方、左侧、右侧和后方60度弧内的最小传感器读数。
- 第三个数据集仅包含前方和左侧的简化距离以及相应的类别标签l。

- 值得注意的是，24个超声波读数和简化距离是在同一时间步长内收集的，因此每个文件具有相同的行数（每行代表一个采样时间步长）。

- 墙壁跟随任务和数据收集旨在检验这一看似简单的导航任务实际上是一个非线性可分分类任务的假设。因此，如感知器网络之类的线性分类器无法学习任务并使机器人绕室无碰撞地运行。能够学习任务并能成功指挥机器人无碰撞运行的，是非线性神经网络分类器，如MLP网络。

- 如果为神经网络分类器提供某种短期记忆机制，通常可以提升其性能。例如，如果将过去输入与当前传感器读数一同提供，即使感知器也能够学习任务并成功指挥机器人。如果使用如Elman网络之类的循环神经网络来学习任务，所得到的动态分类器能够使用比MLP网络更少的隐藏神经元来学习任务。

- 构建具有不同数量传感器读数的文件，旨在评估分类器对输入数量的性能影响。

### 内容

File sensor_readings_24.csv:
- US1：位于机器人前方的超声波传感器（参考角度：180°）-（数值：实数）
- US2：超声波读数（参考角度：-165°）-（数值：实数）
- US3：超声波读数（参考角度：-150°）-（数值：实数）
- US4：超声波读数（参考角度：-135°）-（数值：实数）
- US5：超声波读数（参考角度：-120°）-（数值：实数）
- US6：超声波读数（参考角度：-105°）-（数值：实数）
- US7：超声波读数（参考角度：-90°）-（数值：实数）
- US8：超声波读数（参考角度：-75°）-（数值：实数）
- US9：超声波读数（参考角度：-60°）-（数值：实数）
- US10：超声波读数（参考角度：-45°）-（数值：实数）
- US11：超声波读数（参考角度：-30°）-（数值：实数）
- US12：超声波读数（参考角度：-15°）-（数值：实数）
- US13：位于机器人后方的超声波传感器读数（参考角度：0°）-（数值：实数）
- US14：超声波读数（参考角度：15°）-（数值：实数）
- US15：超声波读数（参考角度：30°）-（数值：实数）
- US16：超声波读数（参考角度：45°）-（数值：实数）
- US17：超声波读数（参考角度：60°）-（数值：实数）
- US18：超声波读数（参考角度：75°）-（数值：实数）
- US19：超声波读数（参考角度：90°）-（数值：实数）
- US20：超声波读数（参考角度：105°）-（数值：实数）
- US21：超声波读数（参考角度：120°）-（数值：实数）
- US22：超声波读数（参考角度：135°）-（数值：实数）
- US23：超声波读数（参考角度：150°）-（数值：实数）
- US24：超声波读数（参考角度：165°）-（数值：实数）
- 类别：前进、轻微右转、急速右转、轻微左转

File: sensor_readings_4.csv:
- SD_front：位于机器人前方60度弧内的最小传感器读数 -（数值：实数）
- SD_left：位于机器人左侧60度弧内的最小传感器读数 -（数值：实数）
- SD_right：位于机器人右侧60度弧内的最小传感器读数 -（数值：实数）
- SD_back：位于机器人后方60度弧内的最小传感器读数 -（数值：实数）
- 类别：前进、轻微右转、急速右转、轻微左转

File: sensor_readings_2.csv:
- SD_front：位于机器人前方60度弧内的最小传感器读数 -（数值：实数）
- SD_left：位于机器人左侧60度弧内的最小传感器读数 -（数值：实数）
- 类别：前进、轻微右转、急速右转、轻微左转

### 致谢

这些数据集从UCI机器学习知识库下载。
Lichman, M. (2013). UCI Machine Learning Repository [http://archive.ics.uci.edu/ml]. Irvine, CA: University of California, School of Information and Computer Science.

### 启发

利用这些超声波读数来预测类别，即给定这些读数，机器人是在直线运动还是在转弯？