elkhawajah / innav

Indoor Navigation Sandbox. Please read README before trying out demo. Demo:

Indoor navigation sandbox

This is a web application to simulate indoor navigation schemes with beacons on a graph-based system. Features include node graph editing, graph path-finding & navigation, signal & sensor simulation, locationing simulation.

This tool is very prototypic, so please open your browser console to see more messages displayed by the sandbox.

Table of Contents

• Features
• Methods of simulation
• File format

Features

The sandbox works by enabling once at a time of the various modes to either edit the graph or simulate a navigation.

• Editing:
  • v or p: add a new node.
    • Physical node is used to represent actual hardware nodes, where virtual node is for arbitrary nodes of the graph.
    • Left-click to place a node on the graph.
  • e: add a new edge.
    • Left-click on two nodes subsequently to create an edge.
    • You cannot create a loop by clicking the same node twice.
  • x: delete an element.
    • Left-click to delete either a node or an edge.
  • m: move a node.
    • Left-click on a node to pick it up, and left-click again to drop it on the new location.
• Simulation:
  • f: find the shortest path between two nodes.
    • No path is created if the nodes are disconnected.
    • Left-click on two nodes subsequently to find the path.
  • l: locate the user.
    • User actual is displayed as bigger red triangle.
    • User accuracy is displayed as smaller orange triangle.
    • Use WASD to move user actual.
    • Use QR to rotate user actual.
  • n: on top of locationing, it finds the path to a node and displays the navigation w.r.t the user orientation.
    • Left-click on a node to set it as destination. Clicking other parts of the graph will set the nearest node as destination.
• c: turn off all modes
  • This clears out the mouse from all modes.
  • It will not stop an ongoing simulation.

Methods of simulation

• Simulating signals
  • There are no actual signal emitters. In other words, the sensoring is not event-based. It is not necessary to have an event-based model for simulating signal emitting and sensoring since we want to keep the sampling at its own pace, without generating useless data in between each sampling.
  • The sandbox simply keeps the readings of sources for later calculation.
  • The sandbox has a scheduled task that counts the distances of all signal sources except those outside a certain bound. The bound is arbitrary, but should be the distance where beacon signal are unreceivable (in reality).
  • The task will then add noise to the distance "actual", which is generated by a gaussian function. The variance of the function is determined by the x-axis inverse of intensity fall-off function in "inverse-squared" form. As a result, the further of a signal source, the larger the noise distribution, hence the less the accuracy.
• Simulating processing readings:
  • The function of noise cancelling simply returns the median of the readings of each signal source.
• Simulating locationing: the sandbox finds the user location by multi-lateration of the stored readings
  • If only one reading: out of luck.
  • If two readings: the sandbox will notify the user to move a little bit to determine the third vector. Locationing with only two readings is badly inaccurate, but won't be way off.
  • If 3+ readings: pick the nearest 3 and trilaterate.
• Navigating:
  • The sandbox shows navigation based on collected user accuracy and calculated path.
  • User accuracy is assumed to be a unit vector, where angle is the user orientation. Path vector is simply the edge to the next node with a direction.
  • The navigation, w.r.t the user, is then determined by subtracting the two vectors.

File format

The application saves to and reads in a simple text file in JSON format. The structure of the object is as following:

{
	"Building":{
		"name":"Awesome building",
		"id":"0",
		"UID":"1"
		},
	"Floors":[
		{
			"id":"0",
			"name":"Good floor",
			"alpha":0,
			"scale":[10,1]
		},
		...
	],
	"Nodes":[
		{
			"GID":"0",
			"Coords":[90,100],
			"Vectors":["001"],
			"Type":"PHYSICAL_NODE",
			"PID":"0",
			"Floor":"0"
		},
		...
	]
}