Incorporating Spatial Constraints into a Bayesian Tracking Framework for Improved Localisation in Agricultural Environments

Abstract

Global navigation satellite system (GNSS) has been considered as a panacea for positioning and tracking since the last decade. However, it suffers from severe limitations in terms of accuracy, particularly in highly cluttered and indoor environments. Though real-time kinematics (RTK) supported GNSS promises extremely accurate localisation, employing such services are expensive, fail in occluded environments and are unavailable in areas where cellular base stations are not accessible. It is, therefore, necessary that the GNSS data is to be filtered if high accuracy is required. Thus, this article presents a GNSS-based particle filter that exploits the spatial constraints imposed by the environment. In the proposed setup, the state prediction of the sample set follows a restricted motion according to the topological map of the environment. This results in the transition of the samples getting confined between specific discrete points, called the topological nodes, defined by a topological map. This is followed by a refinement stage where the full set of predicted samples goes through weighting and resampling, where the weight is proportional to the predicted particle's proximity with the GNSS measurement. Thus, a discrete space continuous-time Bayesian filter is proposed, called the Topological Particle Filter (TPF).The proposed TPF is put to test by localising and tracking fruit pickers inside polytunnels. Fruit pickers inside polytunnels can only follow specific paths according to the topology of the tunnel. These paths are defined in the topological map of the polytunnels and are fed to TPF to tracks fruit pickers. Extensive datasets are collected to demonstrate the improved discrete tracking of strawberry pickers inside polytunnels thanks to the exploitation of the environmental constraints.

Navigate-and-Seek: A Robotics Framework for People Localization in Agricultural Environments

“…Here, the adoption of a GNSS-based system still proves to be the most effective one. In [1], a particle filter is proposed for localizing humans over a topological map. Because it relies only on the GNSS signal, the performance is limited by the bias affecting the GNSS signal itself.…”

“…The method also assumes a fixed velocity model of the humans' dynamics which does not scale well when people move with different speed and directions. In the current work, we build from [1] attempting to solve the above issues by allowing to integrate multiple sensor modalities (rather than only GNSS), estimating the velocity of targets alongside their position, and implementing a mechanism to allow particles to recover from a wrong initialization.…”

“…The Topological Particle Filter (TPF), introduced in [1] and extended in this work, is a method for tracking the position of targets on a map that exploits the structure of the environment -the topological map -to find the closest node to the targets. The distribution of particles P over the topology is used to approximate the probability distribution of the targets on the map.…”

“…Such sparsely connected topological maps allow to constrain the movements of hypotheses along the defined topological edges (see Fig. 2 (bottom)), in order to provide a more accurate overall localization estimate, compensating for substantial noise of sensors, as first proposed by Khan et al [1]. It shall be noted that the benefits of such an approach 1 are not limited to agricultural robotics, but can be utilized in any tracking application that features a sparsely connected topological map.…”

“…The initial work [1] solved the problem employing a "Topological Particle Filter" (TPF) and focused entirely on exploiting such sparsely linked topological graphs to compensate for the noise and inaccuracies of deployed cheap GNSS IoT sensors or mobile phones carried by human workers, to allow them to summon a robot to their locations as part of a robotic in-field logistics solution. This paper takes this idea and the TPF approach further, making the following tangible contributions: (i) Generalization of the TPF framework to support a larger variety of information sources to be fused, i.e., complementing noise GNSS information with detections of humans from a robot's onboard sensing such as LIDAR-based and RFID identification; (ii) complementing the opportunistic and passive sensing approach proposed in [1] with an intelligent decision-making active sensing approach known as Next-Best-Sense (NBS) [2] to utilize the onboard sensors of the robotic platforms to further reduce the uncertainty of humans' location in the field, i.e., to actively move robots around the environment to systematically reduce sensing uncertainty; (iii) a comprehensive evaluation of the proposed integrated approach in a digital twin of the real-world logistics scenario, in which a fleet of robots is supporting human fruit pickers in soft-fruit production, by allowing them to focus on the picking of fruits themselves, while the robotic fleet automates the transportation of picked fruits from the human picker to a local storage facility.…”