Problem

A large problem in the logistics industry is the need for manual inventory management. People often have to count inventory in large warehouses to accurately keep track of items to reduce lost inventory. These manual counting processes are slow and inefficient. Therefore XOsight deploys autonomous drone technology to make the inventory counting process faster and more efficient, while reducing associated costs. 

Counting operations can involve scanning of full pallets by reading their label with a computer vision method. However in many cases also individual boxes on partially picked pallets need to be counted as well. Manually counting these pallets is performed by observing the stack of boxes from a certain perspective and making assumptions of the box size, shape and stacking order to estimate the total amount of boxes, including the ones obstructed from view. 

Current development at XOsight is focused on performing automatic counting by processing a sequence of images gathered by a drone or vehicle based camera. The basic system is currently being proven in real warehouse environments. The current assignment will support this developments in the following ways: Implementing a simulation environment for efficient testing, testing an improved method for visual 3D reconstruction using deep learning and improving the current AI based object understanding methods.

 Proposed approach

• Selecting an optimal simulation environment and setting it up
• Benchmarking current methods
• Literature survey into recent deep learning methods for 3D scene reconstruction and understanding
• Implementing and validating the new methods in simulation
• Validating the new methods on real data.
• Documenting and project handover

XOsight B.V. is part of the YESDelft! ecosystem and is situated in the RoboValley building at the Julianalaan in Delft.

Problem

A large problem in the logistics industry is the need for manual inventory management. People often have to count inventory in large warehouses to accurately keep track of items to reduce lost inventory. These manual counting processes are slow and inefficient. Therefore XOsight develops autonomous drones for automated inventory counting in large warehouses using Artificial Intelligence. Thanks to our cutting-edge customized software, these drones can autonomously navigate and detect inventory in GPS-denied environments, saving companies valuable time and costs on manual inventory counting.

Our drones fly autonomously without direct control inputs. While flying they need to keep an appropriate distance to surrounding obstacles at all time. These obstacles are being detected by analyzing the images of multiple on-board cameras. In the current assignment the collision avoidance capabilities of our system have to be improved by enhancing both the perception as well as the planning pipeline with new methods. Obstacle detection can be improved by employing novel deep learning-based depth sensing. Furthermore path planning can be improved by implementing recently available stat-of-the art method for energy and speed optimized flight.  

Proposed approach

• Assessment of current system collision avoidance
• Review of the state-of-the-art in depth perception and path planning methods
• Implementing improved methods in simulation
• Testing on the real drone system
• Documenting and project handover

XOsight B.V. is part of the YESDelft! ecosystem and is situated in the RoboValley building at the Julianalaan in Delft.

Problem

A large problem in the logistics industry is the need for manual inventory management. People often have to count inventory in large warehouses to accurately keep track of items to reduce lost inventory. These manual counting processes are slow and inefficient. Therefore XOsight develops autonomous drones for automated inventory counting in large warehouses using Artificial Intelligence. Thanks to our cutting-edge customized software, these drones can autonomously navigate and detect inventory in GPS-denied environments, saving companies valuable time and costs on manual inventory counting.

Our drones fly autonomously without direct control inputs. While flying they need to keep an appropriate distance to surrounding obstacles at all time. These obstacles are being detected by analyzing the images of multiple on-board cameras. In the current assignment the collision avoidance capabilities of our system have to be improved by enhancing both the perception as well as the planning pipeline with new methods. Obstacle detection can be improved by employing novel deep learning-based depth sensing. Furthermore path planning can be improved by implementing recently available stat-of-the art method for energy and speed optimized flight.  

Proposed approach

• Assessment of current system collision avoidance
• Review of the state-of-the-art in depth perception and path planning methods
• Implementing improved methods in simulation
• Testing on the real drone system
• Documenting and project handover

XOsight B.V. is part of the YESDelft! ecosystem and is situated in the RoboValley building at the Julianalaan in Delft.

XOsight is start-up which develops tech for automated inventory counting in large warehouses using drones and AI. Thanks to our cutting-edge customized software, these drones can autonomously navigate and detect inventory in GPS-denied environments, saving companies valuable time and costs on manual inventory counting. We cooperate with selected companies to demonstrate the benefits of our technology in real-world operations.

We are looking for an intern who can assist us in the area of business development and marketing. The position offers the possibility to unleash your commercial and creative talents and leave a lasting mark on the organization. 

Your tasks

• Communicating with existing contacts and clients
• Identifying and reaching out to potential clients
• Strategic development and partnerships
• Arranging online and offline marketing materials
• Help prepare for demonstrations, events and tradeshows

Requirements

• Following a business orientated bachelor or master study program
• Excellent communication skills in English, Dutch is a benefit
• Minimal availability 32 hours per week
• Availability during the first half of 2021

What we offer

• Dynamic startup culture
• Possibility to work from home
• Responsibility and the chance to make a lasting impact on our organization

Background

When you buy something online your package will first pass through a large warehouse storage facility. In order to ensure a same day delivery commitment it is of critical importance to keep an accurate inventory account. However due to the large scale of the operation and the human component result in a significant number of inventory errors. To reduce these errors it is common practice to perform manual inventory counting as verification. As one can probably imagine this is a very labor intensive and time consuming task with high costs. Using autonomous drones instead allows for increased frequency of inventory checks at lower costs and higher certainty of inventory level as well as better visibility to out-of-stock and excess inventory. Currently, our product is at the stage of a minimum viable, product. We are part of the YesDelft! accelerator and the RoboValley Ecosystem in Delft.