CJ Logistics

Behind the fast and accurate delivery services we rely on every day, vast amounts of data and sophisticated algorithms are quietly at work. The logistics industry is rapidly transforming into a technology-intensive field driven by artificial intelligence (AI) and robotics. At the forefront of this paradigm shift, CJ Logistics continues to explore diverse ways to quickly apply innovative IT technologies to real-world logistics operations. Among these efforts, the Future Technology Challenge, a hands-on logistics technology competition, goes beyond simply proposing ideas. Participants are challenged to tackle real operational problems by developing and coding practical solutions themselves. Now in its sixth year, the event has firmly established itself as a platform for discovering AI talent capable of leading the future of logistics, once again concluding successfully this year.


Record-High Competition and a One-Stop Platform Creating an Immersive Experience This year’s competition drew greater enthusiasm than ever before. Applications were accepted for about a month from mid-May, attracting 338 participants from 208 teams and setting a new record with approximately 104 teams competing for each challenge. During the main challenge period, which began in June, 98 teams completed their assigned tasks, also marking the highest participation rate in the competition’s history.

As the competition serves as a strategic platform for identifying top talent from leading universities in Korea and abroad, this year’s field attracted an especially strong group of highly skilled participants. A multi-channel promotional strategy played an important role, ranging from prominent advertising on competition platforms popular among university students to active engagement through CJ Logistics’ student ambassadors. One especially notable point was the return of 39 participants from previous competitions. With an overall satisfaction score of 4.3 out of 5 and 93% expressing an intention to participate again, these figures clearly demonstrate how much the competition contributes to participants’ growth.
 

A key factor enabling so many participants to immerse themselves deeply in the challenges was the significantly enhanced in-house online platform. The organizers built a fully integrated One-Stop system covering the entire process within a single environment, from application and guideline review to code submission, real-time scoring, and leaderboard rankings. With fair evaluations conducted using Hidden Data, participants could monitor score gaps with competing teams in real time and continuously refine their algorithms. Over the four-week period, this process fostered a healthy engineering culture focused not merely on chasing scores, but on independently improving and advancing code.

[Challenge 1] Determine the 3D Dimensions of Overlapping Cargo Using a Single CCTV Camera  Accurately identifying cargo dimensions (length, width, and height) at logistics centers is one of the most fundamental yet critical tasks for optimizing vehicle allocation by shipment volume, analyzing loading efficiency with precision, and ultimately reducing substantial operating costs. The first challenge, “CCTV Video-Based Cargo Object Analysis,” was a highly demanding mission requiring participants to accurately estimate the three-dimensional size of boxes moving along a high-speed conveyor belt using footage from only a single CCTV camera. They had to overcome numerous obstacles through technology alone, including deep shadows caused by lighting, occlusion between boxes, and changes in conveyor speed.
 

The grand prize in this challenging category went to a team from Kyung Hee University, consisting of Lee Chan, Lim Juhyeon, and Oh Yunjin. The team first isolated only the regions necessary for analysis, removing unnecessary backgrounds and noise. They then presented an original four-channel deep learning ensemble technique, combining multiple AI models to analyze and integrate information on object location, size, distance, and shape in stages. As a result, they achieved an outstanding average error of just 16.9 cm in the final hidden-data test, earning high praise from the judges. Previously, cargo dimensions could only be measured when each item passed individually through an expensive Intelligent Scanner (ITS) line, often creating bottlenecks. If the camera-based technology identified through this competition is successfully deployed in actual operations, it could significantly improve truck loading efficiency without requiring additional large-scale equipment, creating considerable potential for industrial impact.

[Challenge 2] Perfect Tetris with Boxes of All Shapes and Sizes, Random Palletizing The second challenge focused on developing a “Random Palletizing Algorithm” to make the most efficient use of storage space at automated logistics hubs and maximize cargo loading efficiency. The key objective was to find a way to stack irregular boxes of different sizes and weights as tightly and securely as possible on the limited space of a pallet, even when the boxes arrived in an unpredictable order. The evaluation was based on a total score of 120 points, combining the loading rate, calculated as the ratio of the actual box volume to the total pallet volume (up to 100 points), with additional points for buffer usage (up to 20 points). In particular, a full 20 points could only be earned by using no buffer (waiting space) at all, making robust algorithm design essential for identifying candidate spaces in real time and repeatedly validating each placement.
 

The final victory in this intense competition went to the UOS-RoboStack team from the University of Seoul, consisting of Hong Suhyeon, Kim Yeonjae, and Lee Yumin. Rather than developing one-off code optimized only for a specific test set, the team focused all its efforts on improving generalization (Generalization) so performance would remain stable even under unexpected conditions or unusual box combinations. Their hybrid strategy combined a reinforcement learning (PCT/RL) model to prioritize candidate loading spaces with physics-based heuristics focused on stability, repeatedly checking for boundary violations and collisions with previously loaded cargo. This approach provided a solid foundation for their dramatic comeback. With a final score of 85.54 points, the team secured first place. Their victory was especially meaningful as they successfully carried on the efforts of senior members from the same laboratory who had continued participating since the second edition of the competition, ultimately achieving the lab’s first-ever championship.

A Hub of Innovation Shaping the Future Through Shared Growth, CJ Logistics’ Vision 

The Future Technology Challenge has grown beyond a one-off event into a major platform for mutual growth, bringing together outstanding young talent and CJ Logistics. In addition to the KRW 20 million prize awarded to the grand prize team, all top-performing participants receive exceptional recruitment benefits, including exemptions from the document screening and first-round interview stages when applying to CJ Logistics in the future. Since the inaugural competition in 2021, a total of 29 award-winning participants have joined the CJ Logistics TES Logistics Technology Institute as full-time employees, demonstrating the lasting value of the competition through real career opportunities. Inspired by the passion of this year’s participants, who overcame seemingly impossible challenges through relentless determination and continuous algorithm improvement, CJ Logistics will continue creating opportunities for new technologies and ideas discovered through the challenge to drive meaningful change in real-world logistics operations.